Probabilistic Reconciliation Analysis for Genes and Pseudogenes M OWAIS MAHMUDI Doctoral Thesis

Probabilistic Reconciliation Analysis for Genes and Pseudogenes M OWAIS MAHMUDI Doctoral Thesis
Probabilistic Reconciliation Analysis
for Genes and Pseudogenes
M OWAIS MAHMUDI
Doctoral Thesis
Stockholm, Sweden 2015
TRITA-CSC-A 2015:03
ISSN-1653-5723
ISRN-KTH/CSC/A-15:03
ISBN 978-91-7595-488-2
KTH School of Computer Science
and Communication
SE-100 44 Stockholm
SWEDEN
Akademisk avhandling som med tillstånd av Kungl Tekniska högskolan framlägges till offentlig granskning för avläggande av teknologie doktorsexamen i datalogi
onsdagen den 15 april 2015 klockan 09.00 i Air, SciLifeLab, Tomtebodavägen 23A,
Stockholm.
© M Owais Mahmudi, March 2015
Tryck: Universitetsservice US AB
To my family
iii
iv
Abstract
Phylogeneticists have studied the evolution of life from single celled organisms to the astonishing biodiversity around us for a long time now. The
relationship between species is often expressed as a binary tree - the tree of
life. Availability of fully sequenced genomes across species provides us the
opportunity to investigate and understand the evolutionary processes, and to
reconstruct the gene and species phylogeny in greater detail and more accurately. However, the effect of interacting evolutionary processes, such as gene
duplications, gene losses, pseudogenizations, and lateral gene transfers, makes
the inference of gene phylogenies challenging.
In this thesis, probabilistic Bayesian methods are introduced to infer gene
phylogenies in the guidance of species phylogeny. The distinguishing feature
of this work from the earlier reconciliation-based methods is that evolutionary
events are mapped to detailed time intervals on the evolutionary time-scale.
The proposed probabilistic approach reconciles the evolutionary events to the
species phylogeny by integrating gene duplications, gene losses, lateral gene
transfers and sequence evolution under a relaxed molecular clock. Genomewide gene families for vertebrates and prokaryotes are analyzed using this
approach that provides interesting insight into the evolutionary processes.
Finally, a probabilistic model is introduced that models evolution of genes
and pseudogenes simultaneously. The model incorporates birth-death process according to which genes are duplicated, pseudogenized and lost under
a sequence evolution model with a relaxed molecular clock. To model the
evolutionary scenarios realistically, the model employs two different sequence
evolution models for the evolution of genes and pseudogenes. The reconciliation of evolutionary events to the species phylogenies enable us to infer
the evolutionary scenario with a higher resolution. Some subfamilies of two
interesting gene superfamilies, i.e. olfactory receptors and zinc fingers, are
analyzed using this approach, which provides interesting insights.
v
Sammanfattning
Biologer har under en lång tid studerat livets utveckling från encelliga organismer till den häpnadsväckande mångfalden omkring oss. Förhållandet mellan
arter uttrycks ofta som ett träd. Tillgång till fullt sekvenserade genom ger oss
idag möjlighet att undersöka och förstå evolutionära processer bättre, samt
att rekonstruera gen- och artträd i mer detalj. Men evolutionära processer,
såsom genduplikationer, genförluster, pseudogenbildning och lateral genöverföringar, gör rekonstruktion av genträd utmanande.
Denna avhandling handlar delvis om två probabilistiska Bayesianska metoder som är utvidgningar av metoder för att hitta genträd med ledning av
artträd. Det utmärkande för dessa, jämfört med tidigare metoder, är att de
är probabilistiska samt att de evolutionära händelserna avbildas på artträdet.
Den modell som används är en integrerad probabilistisk modell som innehåller
genduplikationer, genförluster, laterala genöverföringar samt sekvensevolution
utan molekylär klocka. Med hjälp av denna metod analyseras genfamiljer från
ryggradsdjur och prokaryoter, vilket ger intressant inblickar i var, i arträden,
genduplikationer och laterala genöverföringar har skett.
Slutligen presenterar vi en probabilistisk modell som både beskriver evolution av gener och pseudogener. Modellen innehåller en födelse-döds liknande
process enligt vilken gener dupliceras, förloras, eller blir pseudogener samt
två modeller av sekvensevolution utan molekylär klocka. För att modellera
de evolutionära scenarierna realistiskt, innehåller modellen två olika modeller för sekvensevolution, en för evolution av gener och en för pseudogener.
Vi analyserar flera underfamiljer av två intressanta stora generfamiljer, dels
luktreceptorer dels zinkfingrar, med hjälp av denna metod. Resultaten indikerar att visa pseudogener kan ha funktion, exempelvis som RNA kodande
gener.
Contents
Contents
vi
Acknowledgements
1
List of Publications
3
1 Introduction
1.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Current Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5
5
5
2 Evolutionary Events
2.1 Historical Perspective . . .
2.2 Micro-Evolutionary Events
2.3 Macro-Evolutionary Events
2.4 Interesting Gene Families .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
7
9
11
15
3 Modeling Evolution
3.1 Models of Sequence Evolution . . . . .
3.2 Distance-Based Methods . . . . . . . .
3.3 Parsimony-Based Methods . . . . . . .
3.4 Maximum-Likelihood-Based Methods .
3.5 Bayesian Methods . . . . . . . . . . .
3.6 Reconciliation-based Approaches . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19
21
23
25
25
26
27
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4 Present Investigations
39
5 Discussion & Conclusion
41
Bibliography
45
vi
Acknowledgements
A journey filled with reflection, excitement, memories, and a few stressful periods finally comes
to an end. This journey enabled me to learn new scientific fields as well as provided me with
opportunities to reflect upon many other aspects of life. I am grateful to all those who supported
me during this period.
Jens Lagergren, my supervisor, supported me the most during this period. He is a brilliant
researcher and a true inspiration for me. His ingenious ideas and lively personality made this
journey very interesting. Jens, I am grateful for both the personal and professional support that
you have extended throughout this period.
I am also thankful to my co-supervisor Lars Arvestad, who is an exceptionally good teacher and
a researcher, for helping me to explore the exciting area of bioinformatics. I am also grateful to
Bengt Sennblad, my co-supervisor, whose elaborate explanations made this journey very exciting.
His clarity of thinking and ability to explain complex ideas in an easy way always impressed me.
I was fortunate enough to start this journey in the presence of guys like Hossein - cancer & opera
specialist, and Joel - my most recent ancestor in the group. I must mention all the comrades of
our alliance; it was a pleasure working with them. I would like to thank Mehmood - a companion
all the way from Pakistan, Hashim - a true family man, and Auwn - a self-made man, for all
the valuable discussions and collaborations. Mattias and Kristoffer, my desk and gym fellows, I
will always envy their physical and mental fitness. I should also thank Pekka for all interesting
puzzles, explanations, and interesting table hockey games. I will always remember Erik for his
ever-smiling personality, technical help and folk music discussions. Lumi and Viktor, my desk
neighbors, thanks for all the fun stories and sci-fi discussions! I must mention all other friends at
Scilifelab, which I was able to meet during this period. This includes, but is not limited to Linus,
Lukas, Jose, Samuel, Ilgar, Arman, Karolis, Ino, Jorrit, Matthews, Wenjing, and Yrin.
This PhD would have been much harder without Ikram - my brother in arms, whose discussions,
jokes, ideas and suggestions made this period really enjoyable. It was an amazing friendship all
the way from the first lecture of ‘Introduction to Computing’ till this time!
Finally, I would like to thank all my family, specially my parents for providing such a strong
support throughout my life. I love you! I would like to thank my wife Lubaina, for bearing all
my stress and providing the critical support throughout this period. Life would have been much
boring without you!
1
List of Publications
I: Mahmudi O, Sjöstrand J, Sennblad B, & Lagergren J.
Genome-wide probabilistic reconciliation analysis across vertebrates.
BMC Bioinformatics 2013; 14(Suppl 15):S10.
II: Mahmudi O., Sennblad B., Arvestad L., Nowick K., & Lagergren J.
Evolution of Pseudogenes: A probabilistic approach.
Manuscript.
III: Khan MA.∗ , Mahmudi O.∗ , Ullah I., Arvestad L., & Lagergren J.
Probabilistic inference of lateral gene transfer events.
Manuscript.
∗
Contributed equally to this manuscript.
3
Chapter 1
Introduction
1.1
Fundamentals
Life propagates itself by the process of reproduction. This process is, however,
not perfect, which leads to a variety of forms of life. Thus the forms of life that
are best suited to its environment, and able to reproduce faster are more likely to
survive in the long term. The suitability of a life form to its environment is also
dependent on the suitability of other life forms interacting with its environment.
The “survival of the fittest” is therefore, a central idea of the process of natural
selection. This simple but powerful idea has enabled life consisting of single cells to
evolve into this amazing variety of forms of life today. Evolution takes place in every
reproductive cycle of an organism, but in terms of the evolution of new species, it
typically takes millions of years. The first life on earth is estimated to have evolved
around 3.6 billion years ago, while the first multicellular organisms are estimated to
have evolved around 2.1 billion years ago. Around 540 million years ago, complex
multicellular organisms evolved. This followed Cambrian Explosion that resulted
in appearance of most major animal phyla that resembles today’s forms of life.
Evolution of life is central to all aspects of modern biology. As the evolutionary
biologist Theodosius Dobzhansky summarized it in 1973, “nothing in biology makes
sense except in the light of evolution”. Studying evolutionary relationships between
genomes across the species helps us developing a better understanding of the human
genome.
1.2
Current Work
This thesis is an attempt to further extend computational methods to reconstruct
the evolutionary history using the major evolutionary processes. In the methods
proposed in this thesis, genomic information of extant species is combined with the
information available about the evolution of species from other sources to get a
clearer picture of the evolution of life. The studies included in this thesis mainly
5
6
CHAPTER 1. INTRODUCTION
attempts to model microevolutionary events such as nucleotide/amino-acid substitutions and macroevolutionary events such as gene duplications, pseudogenizations,
gene losses, and lateral gene transfers. The methods proposed in this thesis are focused on obtaining a detailed picture of evolution in the sense that the evolutionary
events of gene families are reconciled to the evolutionary events of species along the
evolutionary history.
The methods proposed are Bayesian and employ Markov Chain Monte Carlobased framework for analyses. The distinguishing feature of this work from earlier
reconciliation-based methods is that evolutionary events of gene tree are mapped to
the species tree on the evolutionary time-scale. We present a probabilistic approach
that reconciles the evolutionary events to the species phylogeny by integrating gene
duplications, gene losses, lateral gene transfers and sequence evolution under a
relaxed molecular clock. The method proposed in paper I considered only gene
duplications and gene losses as evolutionary events. Its extension (paper III) is
a more comprehensive method that considers lateral gene transfers as well and is
therefore, more biologically realistic and also applicable when analyzing prokaryotes. Using these methods we analyzed genome-wide gene families for vertebrates
and prokaryotes that provided interesting insights into the evolutionary processes.
The approaches enable us to develop a better understanding of evolutionary processes and can be further used to analyze possible interaction of evolutionary and
ecological processes. Paper II proposes a probabilistic model that models evolution of genes and pseudogenes simultaneously. The model incorporates birth-death
process according to which genes are duplicated, pseudogenized and lost under a sequence evolution model with a relaxed molecular clock. To model the evolutionary
scenarios realistically, the model employs two different sequence evolution models for the evolution of genes and pseudogenes. The reconciliation of evolutionary
events to the species phylogenies enable us to infer the evolutionary scenario with
a higher resolution. Some interesting subfamilies of two interesting gene superfamilies, i.e. olfactory receptors and zinc fingers, are analyzed using this approach,
which provides interesting insights.
The rest of the thesis is organized as follows. A historical perspective of evolution
and a brief introduction to the current understanding of evolutionary processes
is provided in chapter 2. The introduction to the evolutionary processes includes
basic introduction to microevolutionary events as well as macroevolutionary events
followed by introduction to some interesting gene families. Chapter 3 provides a
brief discussion about some of the current methods used to model the evolutionary
processes. Present investigations/studies are described in Chapter 4. Conclusion
and future directions are given in Chapter 5.
Chapter 2
Evolutionary Events
Genes are thought to be the basic functional unit of evolution, which evolve through
various evolutionary processes. Genes with similar characteristics that are believed
to have evolved from a single ancestral gene, are grouped into gene families. Gene
families mostly evolve through microevolutionary events, such as mutations, and
macroevolutionary events, such as gene duplications, gene losses, and lateral gene
transfers. This chapter provides an introduction to the evolutionary processes that
shape genomes. In the following, section 2.1 provides a historical perspective of the
theory of evolution and genetics, which is followed by section 2.2, which contains
a brief description of the microevolutionary processes that shapes the genome of
a species. Macroevolutionary processes that are responsible for the evolution of
genomes will be discussed in the following section (2.3). These macroevolutionary
events include gene duplication events, gene loss events, gene pseudogenization
events and lateral gene transfer events. The last section (2.4) provides introduction
to some interesting gene families that are used in some of the studies of this thesis.
2.1
Historical Perspective
The theory of evolution was popularized mainly by Charles Darwin, but it can be
dated back even further. In attempts to survey the coast of lower South America,
he embarked on HMS beagle in 1831. He spent most of the time investigating
geology and collecting natural history collections. The voyage ended in 1836 but it
was not until 1859 when he published his book, ‘On the origin of species’ [25]. The
book received much attention at that time attracting both positive and negative
criticism.
In his later work, Darwin also explained his Pangenesis theory, according to which
cells in organisms shed tiny particles called gemmules, that circulate throughout
the body and reach the gonads. The acquired fitness of the parents during their life
time is therefore inherited by the offsprings. Francis Galton conducted a wide range
7
8
CHAPTER 2. EVOLUTIONARY EVENTS
of experiments from 1869-1871 in consultation with Darwin to see if gemmules were
circulated in blood. He found no evidence of the characters transmitted through
the transfused blood between dissimilar breeds of rabbits [16].
The inheritance from parents to offsprings was further investigated by Gregor
Mendel, when he conducted a detailed study on peas between 1856-1863, and studied seven characteristics of pea plant, i.e. plant height, pod shape and color, seed
shape and color, and flower position and color. He concluded that traits were inherited in a predictable manner through independent assortment and segregation
of elements. Mendel’s laws of inheritance later on replaced the pangenesis theory
proposed by Darwin [149]. He used the term gene for the first time, when he noticed that the biological variations are inherited from parents to offspring as specific
discrete traits [99].
In 1880s, August Weismann, conducted a set of experiments on mice, removing
the tails of 68 white mice repeatedly for five generations. He reported that no
mice were born in consequence without a tail or even with a shorter tail [150]. His
main contribution was the germ-plasm theory, according to which inheritance only
takes place through germ cells, such as egg cells and sperm cells. Other cells of the
body do not have any role in inheritance. Hugo de Vries made the work of Gregor
Mendel well known, and suggested that Mendelian traits are the heritable elements
transferring along the germ line. He also reconciled the Darwin’s pangenesis theory
with Weismann’s germ-soma cells distinction, by saying that pangenes according
to Darwin theory are located in the nucleus of cell, which when expressed can
move into cytoplasm to change cell structure. By conducting careful studies of wild
variants of Oenothera lamarckiana, Hugo de Vries showed in 1901 that distinct new
forms can suddenly arise in nature, that can last for several generations without
dissipation [29]. These big changes or small changes that suddenly arise in the
offspring, and propagate in the future generations were called ‘mutations’.
In the beginning of twentieth century, genes were discovered as the basic functional
units of heredity that resides on chromosomes, a discovery that was awarded with
Nobel Prize in 1933. In the 1930s, Haldane [62] and Muller [102] hypothesized
that new gene functions may emerge from refashioned copies of old genes based
on cytological observations of chromosomal duplications. This highlighted, for the
first time, the importance of gene duplication in the evolution of genomes. Later
in 1944, Avery et al. [5] showed through experiments that the genetic information
was contained in the DNA of chromosomes. At the end of twentieth century, J. B.
S. Haldane, Sewall Wright, and Ronald Fisher set robust foundations for modern
statistical science, and population genetics. Modern evolutionary synthesis (also
referred to as neo-Darwinian synthesis) was introduced between 1936-1947, integrating all the previous theories. It basically connected natural selection, mutation
theory, Mendelian inheritance to explain patterns observed across the tree of life
using fossils information.
2.2. MICRO-EVOLUTIONARY EVENTS
9
Later on the discovery of DNA structure by James Watson and Francis Crick in 1953
[148] revolutionized the field. Deoxyribonucleic Acid (DNA) is a molecule that is
composed of simpler units called nucleotides. There are four types of nucleotides in
DNA. All the genetic information that is used in the development and functioning of
living organisms is encoded by the sequences of these four nucleotides. The central
dogma of molecular biology was introduced by Francis Crick [23, 24], according to it
information flows from DNA to mRNA, to protein but never from protein to nucleic
acid. The messenger RNA (mRNA) is a ribonucleic acid molecule that transfers
genetic information from DNA to the ribosome, where the triplets of nucleotides
specify the amino-acid sequence for the protein products. This paved the way for
studying hereditary changes from one generation to another.
Discovery of the accurate model of DNA inspired molecular biologists to study the
sequence of DNA, however, it was not possible until 1970s when first Ray Wu and
his colleagues [154], and later Frederick Sanger and Charles Coulson [123] described
methods for determining the sequence of DNA. This opened a new field of molecular
phylogenetics. Sequencing technologies became more and more sophisticated with
the passage of time. Complete genome sequencing projects followed, and the first
complete genome of a free living organism (bacterium) was sequenced in 1995.
Human genome was sequenced by the Human Genome Project in 2003 [19]. This
was followed by the sequencing of around 1092 genomes of different individuals by
the 1000 Genomes Project to study human genetic variation [20]. Genomes of other
species from different clades of the tree of life were also sequenced in the meanwhile.
The development of faster and cheaper sequencing technologies made it possible to
launch projects like Genome 10K Project [67] and 1000 Plant Genomes Project
[153]. The availability of genomic sequences provided the opportunity to discover
the functional as well as non-functional elements of genomes, and it became possible
to compare entire genomes and study the relationships between them.
Understanding how inheritance works at molecular level significantly improved our
knowledge about the relationship of the genetic code to that of physical characteristics. Sequencing the genomes of different species provides the opportunity to
reconstruct the evolutionary history based on the sequence data along with the fossils information. The disjoint facts of our natural history, that were earlier difficult
to understand without the concept of evolution, are now a coherent explanation of
how life evolved on this planet.
2.2
Micro-Evolutionary Events
One of the most basic process of all genetic processes that produce genetic diversity
is a spontaneous point mutation in the germ cells lineage. Mutations in the germ
cells are passed on to the offsprings, while the mutations in somatic cells are specific
10
CHAPTER 2. EVOLUTIONARY EVENTS
to the individual. A point mutation is typically caused by a change of DNA sequence. In a protein-coding DNA sequence, triplets of nucleotides (codons) encode
for a specific amino-acid. A mutation in any of the nucleotides may or may not
result in the change of encoding amino-acid. The mutation may result in a silent
mutation (synonymous mutation), change of amino-acid (non-synonymous mutation), or a frame shift in a sequence (which may be caused by deletion or addition
of a nucleotide).
There are many potential causes due to which a DNA sequence can undergo a point
mutation. Errors during DNA replication is one of the causes of a point mutation.
Another cause is exposure to ultraviolet or high-frequency light ionizing electrons
which in turn may lead to a point mutation. Similarly, reactive oxygen molecules
with free radicals (a byproduct of cell metabolism) can also be harmful for the DNA.
Bonds in DNA also degrade in some cases, due to which keeping the integrity of
DNA becomes difficult.
Spontaneous mutations are context-dependent, i.e. its occurrence depends on the
neighboring nucleotides. Drake et al. found that the mutation rates are different in
germline cells and somatic cells, and the mutation rate in somatic cells was higher
than germline cells for murine [35]. The mutation rate also depends on the size of
the genome and number of germline cell division per generation [87, 35]. A DNA
sequence consists of functional as well as non-functional regions. Around 2% of
the human genome encodes proteins [39], while around 10 percent of the genome
consists of non-protein coding functional regions such as regulatory sequences and
others, that are under significant selective pressure [59]. Much more is known
about the protein coding regions of the human genome than the non-protein coding
functional regions. However, ENCODE was able to associate 80% of the human
genome with at least one biochemical function [21].
Deleterious mutations are those that make the organism less fit in its environment
and are eventually eliminated from the population over the time. Advantageous
mutations contributes towards the fitness of the organism, and are much more likely
to prevail in the species in the long run. If the number of synonymous mutations
per synonymous site (dN ) is larger than the number of non-synonymous mutations
per non-synonymous site (dS), i.e. (dN/dS < 1), it is believed that the sequence
has underwent negative selection or purifying selection. Higher rate ratio of nonsynonymous to synonymous mutations (dN/dS > 1) implies that the sequence has
been under positive selection or diversifying selection. A sequence is thought to
have been evolving neutrally, if the the ratio between the two types of mutations
is approximately equal to one (dN/dS ≈ 1). If a nucleotide changes from a purine
(adenine and guanine) to purine, or from a pyrimidine (cytosine and thymine) to
pyrimidine, it is called a transition. On the other hand, if a nucleotide changes
from a purine to pyrimidine or vice versa, it is called transversion. Transversion
involves much greater change in the structure as compared to transition and are
2.3. MACRO-EVOLUTIONARY EVENTS
11
therefore not as common as transitions. Mutations in any functional part, e.g., the
regulatory regions of the genome can also make an organism less or more fit in its
environment, depending on the change it brings in the gene function.
In the majority of cases living organisms invest heavily in preventing mutations,
e.g. they develop mechanisms, to export or destroy substances causing mutations,
and to repair damaged DNA. In certain environments, such as stress environments,
a higher mutation rate may bring advantageous mutations necessary for the survival of the species [55]. However, deleterious mutations gained become a fitness
cost for the species in the long run, when the stress period is over [53]. Therefore, a balance between deleterious mutations, and rare advantageous mutations
increases the long-term fitness [114]. Extreme environmental changes, and mass
extinctions are also closely associated with the genetic diversity leading to appearance of phenotypic novelties [70, 57, 152]. Although the exact mechanism of how
stress-induced variation occurs in the genome is not clear, it is known that stress
plays an important role in the evolution of genomes [6].
2.3
Macro-Evolutionary Events
In 1970, Susumu Ohno published his famous monograph, in which he suggested
that gene duplication is a mechanism responsible for the origin of most novel genes.
He also introduced two possible fates for the genes originated from a gene duplication. After a gene duplication, one of the two copies of genes can carry out the
ancestral function, while the other gene is free to evolve a new gene function, socalled neo-functionalization. The ancestral gene is more likely to keep the original
function, while the duplicated copy may evolve a new gene function. However, the
most probable fate of a duplicated gene copy is pseudogenization and consequently,
majority of the duplicate genes are lost during evolution. A third possible fate of a
gene duplication, sub-functionalization, was introduced by Stoltzfus [133] and Force
et al. [51]. In this case, both of the child genes keeps on performing the ancestral
function for a while, and eventually specializes in a subset of the functions performed by the ancestral gene. The three fates of gene duplication are also discussed
in article published by Lynch et al. [94], where they conclude that in most cases
gene duplication experience a short period of relaxed selective pressure which ends
with a silencing mutation. In a few cases, the duplicated gene survive long enough
to develop a new function, which is subsequently put under selective pressure.
Estimated rates of gains and losses varies across different clades in the tree of life.
Anciently diverged group of species such as yeasts, mammals and flies have similar
average duplication/loss rates [31]. In mammals the rate of gain and loss of genes
per million year is around 0.0016, in flies the rate is around 0.0012, while in yeast
the rate is around 0.0020 [30, 61]. The rate is higher in the primates where the rate
of gain/loss of genes per million year is 0.0024 [60]. In great apes this rate is even
higher, i.e. 0.0039 gain/loss of genes per million years [60]. Other lineages such as
12
CHAPTER 2. EVOLUTIONARY EVENTS
dog, mouse and rate have low rates (like the overall trend of mammals) i.e. 0.0014
gain/loss of genes per million years. In drosophila the rate of gain/loss of genes
per million years varies between 0.0006-0.0193. However, low coverage of genomes
in some cases is a source of error in the gain/loss estimates. The highest rates are
estimated for genomes having low coverage [61], and consequently those genomes
should be treated with caution.
Gene Duplication
A number of mechanisms may cause a gene to duplicate. These mechanisms include unequal crossing-over, segmental duplication, whole genome duplication, and
retro-transposition. Unequal crossing-over only happens in diploid cells. During
meiosis, the duplicated chromosomes attach to each other at the centromere region. Reciprocal or non-reciprocal recombination may take place during this time.
In the case of unequal crossing-over during recombination, one chromatid receives
a longer segment, while the other receives a shorter one. If there are similar regions, such as repetitive elements, on the chromosomes, unequal crossing over is
more likely to occur. This may result in loss of some genes from one chromosome
and addition of those genes to the other chromosome. Similar genes on one of the
chromosomes may subsequently induce further unequal crossovers. Loss or gain of
genes may result in change of gene dosage. Gene dosage is the number of copies of
a gene present in a cell or nucleus. Increase or decrease in gene dosage may result
in higher or lower levels of gene products provided that it is not regulated elsewise.
Around 30% of all recent human segmental duplication are reported to be caused
by recombination-like mechanism, and more or less similar figure has been observed
for rodents genome [163]. Segmental duplications are found to be large and highly
identical in humans. This has also been observed for the chimpanzee genomes
[17]. The high identity of segmental duplications makes the genome unstable and
more vulnerable to the rearrangements through recombination events. Using wholegenome shotgun sequence detection method, segmental duplications for human were
observed to be 5.32%, chimpanzee 3.78%, orangutan 1.18%, rhesus macaque 1.55%,
mouse 6.13%, dog 0.82%, cow 5.63%, and opossum 2.7% [97].
Also whole genome duplications may happen during evolution, in which the entire
genome is duplicated in the offspring. Ohno hypothesized that the early vertebrates went through two rounds of whole genome duplications [111]. The lineage of
ray-finned fishes is also believed to have undergone a third round of whole genome
duplication [101]. Flowering plants are believed to have frequent whole genome duplications that enabled them to evolve into such an astounding variety [78]. Whole
genome duplication may result in duplication of all genes, which follows a substantial relaxation of selective pressure. Although most of the duplicated genes are
destined to get pseudogenized and lost from the genome ultimately, some of the
2.3. MACRO-EVOLUTIONARY EVENTS
13
genes develop new functionality thus creating diversity in the gene function and
ultimately new phenotypes.
Barbara McClintock, while studying various mutation patterns in maize discovered
that some segments of DNA had the ability to move around in the genome [98].
The ‘jumping genes’ were later on known as transposons (a.k.a. mobile elements or
transposable elements). Transposons have the ability copy themselves to another
location of the genome, or cut-and-paste themselves to the target location. In bacteria, DNA-only transposons are common, which cut-and-paste themselves to the new
genomic location using a transposase enzyme without the involvement of an RNA
intermediate. In the process of retrotransposition (common in eukaryotic cells),
genes are transcribed to RNA (mRNA) and then retrotranscribed back to DNA
(cDNA) using a reverse transcriptase and are then inserted back into some random
positions in the genome. The enzyme necessary for this process, reverse transcriptase, are encoded by different retrotransposable elements in different species, for
example LINE-1 retrotransposons provide the required enzymes in mammals. The
resulting retrotransposon have no introns, and generally reside on a different chromosome. If the retrotransposon lands in an intron of another gene, it is transcribed
with that gene, if it, in contrast, lands outside the existing genes, the most probable
fate is pseudogenization, as it lacks regulatory sequence. In the case a regulatory
sequence is evolved by chance, this may result in addition of a new gene to the
genome also known as retrogenes. Retrogenes which are eventually transcribed are
much more likely to adapt a new gene function compared to what genes duplicated
by other mechanisms are.
Detailed analyses of some hominoid young retrogenes suggest that they have contributed to brain evolution [81]. Sometimes retrogenes, such as Rps23, evolve a
completely new gene function by simply being transcribed in reverse direction in
the new genomic location [159]. Some striking phenotypic changes were also associated to a retrogene derived from a growth factor gene (FGF4). It was reported
that the conserved, expressed retrogene fgf4 contributed towards a morphological
change i.e. short-legged phenotype of several common dog breeds [113]. In this
case the change of phenotypic characteristic is more likely to be associated with
only the change of gene dosage. That is, a gene duplication has the potential to
innovate new phenotypic characteristics by manipulating gene dosage [81].
Lateral Gene Transfer
Lateral gene transfer (LGT) or horizontal gene transfer is the transfer of a gene
from one organism to another organism such that neither of them is ancestor of the
other. Lateral gene transfer is common in bacteria and archae and can be mediated
by viruses, plasmids and transposons [22]. LGT is also believed to be a primary
reason for the evolution of resistance in bacteria against antibiotics [83]. The role
of LGT seems to be limited in the case of metazoans, although some cases have
14
CHAPTER 2. EVOLUTIONARY EVENTS
been reported [161, 84]. Among bacterial genomes, LGT is often observed between
distantly related species [82]. The key mechanisms for LGT are transformation, conjugation and transduction. Transformation is the direct uptake of foreign genetic
material through the cell, conjugation is the transfer of genetic material through
a bridge-like structure between the two cells, while transduction is the insertion of
foreign genetic material through bacteriophages. Various types of mobile elements
are also among the important forces that drives the genomic re-arrangements. Lateral gene transfers challenges the classical definition of species, and the assumption
of tree-like evolution of the species. Since the LGTs are observed between distant
bacterial species, they are a confounding factor in the inference of phylogenetic
trees that are based on the gene sequences.
Process of Pseudogenization
One of the three possible fates of a duplicated gene is pseudogenization. As mentioned in the earlier section, after a gene duplicates, one of the two genes evolves
under relaxed selective pressure for a while, and in most cases it gets ‘pseudogenized’. The pseudogenized genes were initially thought to be nonfunctional genes
and often termed as ‘junk DNA’. Jacq and his colleagues used the term pseudogene
for the first time, when they discovered a version of the gene coding 5S rRNA that
was truncated but retained the homology with the active gene in Xenopus laevis
[77]. They are also frequently referred as ‘genomic fossils’, as they are thought
to be evolving neutrally [162]. Pseudogenes are present in a wide range of species
including prokaryotes [103], plants [10], insects [66], nematode worms [65], but they
are particularly numerous in mammals [160].
A gene may get transformed into a pseudogene in many ways. After a gene duplicates, the selective pressure on the resulting genes is relaxed for sometime and one
of the genes may, therefore, accumulate non-sense mutations. This may result in a
frameshift (deletion or addition of a single nucleotide), thus resulting in disruption
of the open reading frame of a gene. A gene might also gain a pre-mature stop codon
thus preventing it from either transcription or translation. Pseudogenes may exist
in a genome in form of a processed pseudogene that results from the process of retrotransposition. In this case, the transcribed mRNA is reverse transcribed to cDNA
and integrated into a new genomic location. If this newly integrated sequence lacks
its own regulatory sequence, it may be ‘dead on arrival’. This processed pseudogene
also lacks intron. A non-processed pseudogene is a gene that pseudogenizes as a
result of accumulating non-sense mutations. Recent studies have, however, shown
that pseudogenes in some cases are still performing some function. For instance
it was found that synonymous mutations were far more frequent than the nonsynonymous mutations in the Drosophila Est-6 pseudogene suggesting a selective
pressure [8]. In some of the pseudogenes present in chicken, i.e. IglV and IghV, and
in mouse i.e. VH, the number of stop codons in the coding sequence region is far
lower than expected under neutral evolution [121, 124]. It has also been observed
2.4. INTERESTING GENE FAMILIES
15
that some pseudogenes retain conservation across species, for example, during the
analysis of major histocompatibility complex extended class II, two pseudogenes
were found to be homologous to human HIV TAT-specific factor-1-like and zincfinger-like pseudogenes [136]. Svensson et al. [137] identified ancient pseudogenes
common to human and mouse that originated before the speciation split and were
still highly conserved. Another study identified 48 conserved pseudogenes in human, mouse, rate and dog suggesting a potential function [96]. Therefore, in order
to understand how the genome functions, it is important to understand the process
of pseudogenization and the potential role of pseudogenes across organisms.
2.4
Interesting Gene Families
A gene family is a set of genes that shares similar characteristics such as similar
sequence, structure or function. The genes in a gene family are thought to have
duplicated and evolved from a single ancestral gene. Such a gene family is called
a homologous gene family A gene family evolve by undergoing processes such as
duplications, losses, pseudogenization, and lateral gene transfers. During evolution
some genes may change significantly in their characteristics from the existing members of the gene family, e.g., by obtaining a different function. The evolution is
also affecting gene families through large segmental duplications and whole genome
duplications that replicates some or all of the existing genes in the genome. As discussed earlier, duplicate copies of genes relax the selective pressure, and results in
neo-functionalization or sub-functionalization of the gene function of the parental
gene.
Rate of evolution varies across the gene families. Some gene families evolve faster
than others and develop new gene functions. Evolution of new gene families is expected to contribute towards the fitness of an organism in its surroundings. Sensing
the environment is one of the critical requirement for the survival of a species. The
genomes of different species have developed sophisticated senses to sense the surrounding environment during the course of evolution. One such gene family is Olfactory Receptors - the largest gene family in vertebrates, that is targeted towards
sensing odors. Regulation of expression of other genes is another important genome
function. It has been shown that paralogs mostly sub-functionalizes at regulation
level and less frequently through changes in their cellular component, biological
process or molecular interactions, and rarely in biochemical function [147]. The
Zinc Finger gene family, the second largest gene family in vertebrates, regulates
other genes by up- or down-regulating their expression. In the following, I will
briefly describe the above-mentioned gene families, i.e. Zinc Fingers and Olfactory
Receptors, that have had higher rates of duplications, losses, and pseudogenization
in some lineages in the tree of life.
16
CHAPTER 2. EVOLUTIONARY EVENTS
C2H2 - ZINC FINGER
The C2H2 Zinc-Finger (ZNF) family is one of the largest gene families in mammals.
It is the second largest gene family in the human genome, and the largest class of
transcription factors. The ZNF genes encode transcription factors that regulate
gene expression by binding to DNA and RNA using the ZNF motifs. Each ZNF
gene typically encodes a number of ZNF motifs, each consisting of 28 amino-acids.
ZNF genes are grouped into multiple subfamilies based on the presence or absence of
N-terminal effector domain such as Kruppel-associated box (KRAB). Mammalian
genomes have expanded a lot by duplicated these genes (see Figure 2.1) [32]. In the
800 700 600 500 400 300 200 100 0 Human Mouse Chicken Fly Worm Yeast Plant Figure 2.1: Distribution of the C2H2-ZNF genes in different genomes. The total number
of C2H2-ZNF genes in different genomes is shown. There has been a massive expansion in
the number of C2H2-ZNF genes over the course of evolution [32]. Humans (Homo sapiens
[140, 33]) have the highest number of genes as compared to the other species for instance
Mouse (Mus musculus [33]), Chicken (Gallus gallus [142]), Fly (Drosophila melanogaster),
Worm (Caenorhabditis elegans), Yeast (Saccharomyces cerevisiae) and Plant (Arabidopsis
thaliana [93]).
human genome, a large proportion of all the C2H2-ZNF genes (more than 70%) are
found in tandemly clustered gene families, whereas the remaining exist as singleton
genes. These clusters have on average six ZNF genes [140]. Detailed studies of these
clusters show that except from duplication of single genes, duplication of partial
and whole clusters are also responsible for the rapid evolution of C2H2-ZNF gene
family [140, 74, 64]. These genes are thought to be a result of unequal crossing over
or DNA slippage during replication, favored by highly identical repetitive motifs
[140]. The unstable genomic location of these clusters is also thought to be one of
the potential reasons for making this gene family prone to duplications and losses
[109]. Recent studies have reported evidence of positive selection for KRAB-ZNF
genes, which suggest a strong drive for neo-functionalization after gene duplication
2.4. INTERESTING GENE FAMILIES
17
in this family [105, 42, 110]. Pseudogenes are mostly produced by partial gene
duplications, whereas they comprise of approximately half of the KRAB-ZNF loci
[109].
OLFACTORY RECEPTORS
Sensing the environment through different types of odors constituted a big advantage for tetrapods during the process of terrestrial adaptation. The ability to
identify odors was evolved by expansion of a gene family known as Olfactory Receptors (OR). The number of olfactory receptor genes vary from species to species,
but they form the largest gene family in vertebrates. The high frequency of OR
genes in mammals suggests that detecting and identifying odors was one of the
basic requirement for the survival of these species. Species that have other sources
of sensing the environment have lower number of OR genes, e.g. primates have
trichromatic vision and less number of OR genes as compared to the dichromatic
mammalian species. Humans have a higher rate of pseudogenization (around 50%)
and consequently a lower number of functional genes (800) in the OR gene family,
while mice have a lower rate of pseudogenization and consequently a higher number
of total functional genes (1400) with (around 20-25%) [107]. The number of functional functional OR genes in mice is 2.7 times higher than humans. Species such as
toothed whales have had an extremely high rate of pseudogenization of OR genes
and have completely lost the olfactory system. The development of echolocation
system, as well as, the adaptation to the aquatic environment might be the possible reasons for the loss of olfactory system [107]. Platypus have a semi-aquatic life
style and using a sophisticated bill sense they can find prey with their eyes, ears,
and nostrils closed. The number of OR genes (718) in platypus is also comparatively low with a higher rate of pseudogenization (around 50%)[107]. In paper II,
we study a model of evolution that also incorporates pseudogenization and analyze
some sub-families of this interesting gene family.
Chapter 3
Modeling Evolution
In this chapter, I will discuss some traditional as well as recent methods developed
in order to model key biological processes involved in the evolution of gene families.
The evolutionary relationships of gene families and species are called gene phylogenies, and species phylogenies, respectively. Phylogenetics is the sub-field of biology,
in which phylogenies are studied by taking into account the molecular sequences,
as well as information from other sources such as morphological data and fossils
records. The molecular sequences may be in the form of DNA sequences or protein
sequences. Phylogenomics refers to the use of phylogenetic methods using large
stretches of genomes across the species to infer the gene function and evolutionary relationships in multi-gene families [127, 88]. The evolutionary relationships
among the genes, species and their populations are usually represented by binary
trees. Two basic phylogeny problems are resolving gene phylogeny and resolving
species phylogeny. A brief introduction about species phylogeny, and its reconstruction methods is given below, which is followed by the introduction of key concepts
of the gene phylogeny and the methods used to construct gene families. The first
five sections of this chapter contains introduction to classical approaches that attempts to resolve gene phylogenies. It includes an introduction to substitution
models, distance-based methods, parsimony-based methods, maximum likelihoodbased methods, and Bayesian methods. The last section describes some of the
recent methods used to reconcile gene phylogenies and species phylogenies.
Species Phylogeny Problem
Attempts to resolve species phylogeny have been made since Darwin first proposed
the tree-like nature of the evolution of species. Phenotypic evidence was initially
used to resolve the evolutionary relationships among species before the availability
of molecular sequences. However, with the discovery of DNA, it became possible
to compare the molecular sequences of two species to study their evolutionary
19
20
CHAPTER 3. MODELING EVOLUTION
relationship. This revolutionized the field, and later on with the availability of ever
increasing computational resources, it became possible to develop sophisticated
algorithms to address phylogeny problems. Many algorithms have been proposed
to study species phylogeny using molecular sequences and to approximate times
associated with the speciation events. The species phylogeny is usually represented
as a binary tree. The ancestor of all the species in the tree is represented by the
root of the tree, while leaves of the tree represent extant species. Least common
ancestor (LCA) of a set of leaves in a tree is the internal vertex of the tree, which
have no additional leaves as its descendants, except than the set itself, . The internal
vertices of a species tree represent speciation events. A speciation event is an event
in the evolutionary history of a species after which a part of the population of the
species evolve so differently from the rest, such that the two groups are typically no
longer able to interbreed and produce fertile offsprings. However, the distinction
between species is sometimes not very clear as Charles Darwin wrote in his book,
“I was much struck how entirely vague and arbitrary is the distinction between
species and varieties” [26]. A speciation event may happen for several reasons,
some of which are: geographic isolation leading to different selective pressures and
genetic drifts, isolation of the population due to other reasons, or even without
isolation merely due to the genetic drift. Fossils of ancestral species may provide
important information for accurate estimation of the age of a specific lineage in
the tree of life; however, fossils records are not readily available throughout the
evolutionary history. Genomes of the extant species on the other hand provides
molecular sequences that can be used to resolve the species phylogeny and date the
tree of life.
A number of methods have been proposed to reconstruct the species phylogeny
from the molecular sequence data. Some of the methods involves the use of a
‘super-gene’ that is formed by concatenating many genes of different gene families
(typically having one-to-one relationship of genes across species). These supergenes are then used to reconstruct the species phylogeny. However, since the gene
families usually do not have the same number of genes across species, it is difficult
to create such representative super-genes. Recently, methods based on summary
statistics that account for coalescence variance have shown promise to reconstruct
species tree [13]. Gene coalescence has also been modeled in ML and Bayesian
methods recently to infer the most likely species tree. Some recent methods that
attempt to solve this problem are STAR [92], STEAC[92], BEST [91], and BEAST
[68].
Gene Phylogeny Problem
Another interesting and challenging problem is resolving the gene family phylogeny.
The gene family phylogeny is typically represented by a binary tree, where the leaves
represent extant genes and the root represents the ancestral gene from which all
3.1. MODELS OF SEQUENCE EVOLUTION
21
genes originated mainly through duplications. A group of genes, which are similar
in gene sequence and gene function, evolving from the same ancestral gene forms a
homologous gene family. The term homology was coined by Owen [112] for the first
time by referring to the “the same organ in different animals under every variety
of form and function". With the advent of molecular sequences, homology is now
also defined as the existence of shared ancestry between a pair of sequences, genes
or their products such as protein sequences. A homologous relationship may be
orthologous, paralogous, or xenologous as defined by Walter Fitch in his seminal
article [49]. Two extant genes in two species are orthologous if both of them evolved
from a common gene in the least common ancestral species (i.e. least common
ancestor occurs due to a speciation event). Alternatively two genes are paralogous,
if they exist within the same species or arise from different genes in the most recent
common ancestral species. Two genes belonging to two different species are xenologs
if they emerge from a lateral gene transfer event. Paralogous genes that arise from
a whole genome duplication event are termed as ohnologs. A gene tree represents
all the evolutionary events (such as duplications, losses, speciations, and lateral
gene transfers) as internal vertices and all the extant genes as leaves of the tree.
Any lineage of the gene tree that is lost during the process, is usually pruned from
the tree. A gene family phylogeny is sometimes represented by structures other
than trees as well, e.g. phylogenetic networks/graphs are also used to represent the
evolutionary relationships involving lateral gene transfers in certain gene families
[14].
Identifying homologous gene families is often the first step in a phylogenomic analysis. This is however, a non-trivial task. Several methods have been proposed to
construct a homologous gene family. These methods usually use one or many of
the three kinds of information: sequence information, copy-number of gene across
extant species, and location of genes on the corresponding genomes (synteny). Synteny is the condition where a block of genes is found in the same relative position in
the genomes of the two species. These methods usually make use of one or more of
the following techniques: similarity-based searches such as BLAST, gene clustering
techniques, gene-function-based searches, and/or gene synteny [52].
For a given gene family, the gene tree is usually inferred from the gene sequences.
There are a number of traditional approaches to estimate the gene tree from the
gene sequences. A brief introduction to each of the broad categories is given below.
3.1
Models of Sequence Evolution
Genomes of species experience constant molecular evolution through random mutations. Genetic drift and selective pressure are two major forces that shape genomes
of two species differently. A number of molecular/sequence evolution models for
nucleotides as well as amino-acids have been proposed. These models are usually based on a Markov chain model of nucleotide/amino-acids substitution. The
22
CHAPTER 3. MODELING EVOLUTION
parameters used to define the rates at which one nucleotide/amino-acid replaces
another during evolution, differentiate one substitution model from the other. The
sequence evolution models are used to compute the likelihood of a phylogenetic tree
(see 3.4, 3.5). A brief description of commonly used nucleotide substitution models,
amino-acid substitution models, and codon substitution models is given below.
First I discuss four commonly used Markov models for nucleotide substitutions.
JC69 (proposed by Jukes and Cantor [80]) is the simplest of all substitution models, having equal substitution rates between all nucleotides. Motoo Kimura [86]
proposed the assumption of different rates for transitions (nucleotide substitutions
that changes purines to pyrimidines or vice versa) and transversions (nucleotide substitutions between purines or between pyrimidines) in 1980. The model is known
as K80. Later on in 1981, Felsenstein proposed a nucleotide substitution model,
F81[46], in which the assumption of equal nucleotide frequencies was relaxed and
the substitution rate corresponds to the equilibrium frequency of the target nucleotide. In 1985, Hasegawa et al. introduced a model(HKY85) that allowed unequal nucleotide frequencies as well as proposed different rates for transition and
transversions.
Sequence evolution can also be modeled in protein sequences. Protein sequences
consists of amino-acids; in the DNA sequence of the corresponding gene, each
amino-acid is encoded by a codon comprising three nucleotides. The genetic code
in case of protein sequences is degenerate, where 20 amino-acids are encoded by
64 codons. Point mutation in any of the nucleotides of a codon may result either in a synonymous mutation, in which the encoded protein does not change,
or in a non-synonymous mutation, in which the encoded protein is changed. Because of the lower selection pressure on synonymous substitutions, they are likely
to occur at a higher rate than non-synonymous substitutions. For distant species,
nucleotide sequences are therefore, more likely to become saturated (i.e. showing
reverse substitutions) as compared to the amino-acid sequences. Thus evolutionary
pressures at the level of protein sequences cannot be modeled appropriately by nucleotide substitution models. The amino-acid based substitution models not only
model the evolutionary pressures at protein level, but they are also more realistic at
longer evolutionary distances. They typically are based on 20x20 exchangeability
matrix between the 20 possible amino-acids. Different amino-acids have different
biological, chemical, and physical properties, due to which they have different rates
of substitution with other amino-acids [90]. For example, isoleucine and valine are
two amino-acids that have a higher rate of substitution between each other as compared to arginine and aspartate. The amino-acid based substitution models are
traditionally empirical models [28, 79]. The substitution rates in these models are
typically estimated from large collections of molecular sequences.
Studying sequence evolution in the amino-acid sequences hides the details of synonymous substitutions (transitions between codons encoding same amino-acid).
3.2. DISTANCE-BASED METHODS
23
Codon-based models, in the contrast, can better model evolutionary pressures at
different codon positions. Codon-based substitution models enable us to study
the relative rate of synonymous substitutions to that of non-synonymous substitutions, which is an unambiguous indicator of positive natural selection at molecular
level [106]. The codon-based substitution matrices are usually 61x61 (without stop
codons), or 64x64 including three stop codons. This matrix consists of the transition
probabilities from any codon to any other codon. Typical codon-based substitution models take advantage of the nucleotide position in a codon by incorporating
ratio between non-synonymous mutations and synonymous mutations. The ratio
of non-synonymous to synonymous rates ω = dN/dS is used to determine if a
gene is going through positive/adaptive selection, negative/purifying selection or
neutral selection. The transitions to transversion rate ratio is also usually incorporated in the codon-substitution model. Transversions (substitutions of purine
to pyrimidine or vice versa) are usually not as common as transitions (substitution of purine to purine or pyrimidine to pyrimidine), because transversions involve
much greater change in the structure. The exchangeability matrix between codons
is determined by different configurations of parameters such as non-synonymous
to synonymous rate ratio, transition to transversion rate ratio, and equilibrium
frequencies of codons. One of the commonly used codon substitution models was
proposed by Bielawski et al. [11]. The instantaneous substitution rate matrix
between codon i to codon j in this case is given as:


0,
if i and j differ at more than one position in a codon triplet





µπ
,
differ by a synonymous transversion
j

qij = µκπj ,
differ by a synonymous transition


µωπj ,
differ by a nonsynonymous transversion



µκωπ , differ by a nonsynonymous transition
j
where πj is the equilibrium frequency of codon j, µ is the normalizing factor, κ is
the transition/transversion ratio, and ω is non-synonymous to synonymous (dN/dS)
ratio. In paper II, we use the above-mentioned codon substitution matrix modeling
pseudogene evolution, where ω is equal to 1 and transition to stop codon has a nonzero probability. To model gene evolution we use the standard codon substitution
matrix described above, where transition to stop codon is not possible. The codon
equilibrium frequencies are estimated from the genes/pseudogenes sequences.
3.2
Distance-Based Methods
Distance-based methods use the dissimilarity of gene sequences to identify phylogenetic relationship among the genes. Using the pairwise distances between all
the sequences as input, distance-based methods attempts to reconstruct a phylogenetic tree with branch lengths. Pairwise-distances between two sequences can
24
CHAPTER 3. MODELING EVOLUTION
be computed using pairwise alignment methods. A brief description of the pairwise alignment methods and multiple sequence alignment methods is given later in
this section. The pairwise-distances are typically used in one of the agglomerative
clustering methods to construct an estimate of the gene tree. One of the most
widely used distance-based methods is neighbor joining (NJ) [122]. This method
start with a starlike tree, and the pair of taxa having minimum pairwise-distance
is successively joined together, until a fully resolved tree is obtained. It is a fast
and efficient method, which makes it possible to obtain confidence estimates using
bootstrapping. Neighbor joining has been implemented in a variety of settings with
some minor improvements (such as BIONJ [54], RapidNJ [126], and FNJ [41] etc.),
improving its efficiency or its accuracy.
Pairwise alignment methods usually belong to one of the two broad categories, i.e.
Dynamic-Programming-based methods (such as Needleman-Wunsch [104], SmithWaterman [131]) or Word-Search Methods (such as BLAST [2]). Dynamic programming (DP)-based methods usually uses a scoring scheme for a match, a missmatch, opening a gap and extending a gap. DP-based methods are guaranteed
to find the optimal alignment given a scoring function. A good scoring function
in practice is usually based on empirical observations. Word-search-based methods such as BLAST are not guaranteed to give optimal alignment, but are much
more efficient than DP-based methods. A gapped BLAST first identifies all matching words between the sequences, then identifies high-scoring segments above the
given threshold, occurring within the allowed distance from another high-scoring
segments, on the same diagonal. A limited use of DP follows to join the high scoring segments. Word-search-based methods are very useful in searching large-scale
databases. Pairwise methods can be used to obtain a phylogenetic tree. Such a
phylogenetic tree can be then used in aligning multiple sequences (more than two
sequences). A brief introduction to the multiple sequence alignments is given in
the following.
Multiple sequence alignment (MSA) methods attempts to align more than two
gene sequences at a time. This is often the first step in the phylogenetic studies,
which can potentially effect the quality of a study. Getting MSA is computationally
challenging and the common variations of MSA are shown to be NP-hard [40]. The
technique of DP-based alignment is applicable to more than two sequences, but is
rarely used in practice since it is computationally expensive both in terms of time
and space. Progressive alignment methods is another class of MSA, that aligns
the query sequences based on an initial tree such that most similar sequences are
aligned first, followed by less similar sequences until all the sequences have been
aligned. The initial tree is based on pairwise alignments of the gene sequences.
T-Coffee [108] and Clustal [18] are some of the examples of progressive alignment
methods. Another class of MSAs is iterative methods, which work similarly to
the progressive methods, except that it realigns the sequences in each iteration, by
aligning a chosen sequence against the profile of remaining aligned sequences. This
3.3. PARSIMONY-BASED METHODS
25
process is repeated until the there is no further increase in the overall alignment
score. There are a number of other MSA methods that uses different heuristics to
efficiently generate MSA. Some other interesting MSA methods use Hidden Markov
Models (e.g. HMMER [48]), genetic algorthims and simulated annealing.
3.3
Parsimony-Based Methods
Parsimony methods are based on the assumption that evolutionary events are rare
and, hence, the tree with minimum number of evolutionary events is the most
likely scenario. Parsimony was originally developed for use in analyzing discrete
morphological characters, but the same principle was later on extended in the phylogenetics. The evolutionary events are typically gene duplications, gene losses and
gene transfers, or in the form of substitutions in the gene sequence [151]. Maximum parsimony has, however, been debated since long. The relative efficiency
of parsimony-based methods over probabilistic methods made it a tool of choice.
Parsimony-based methods should be used with caution as it has been shown that
they can be misleading in some scenarios [45]. They have been widely used in
the biological community in different settings. Some of the commonly used tools
include MEGA5 [141] and PAUP [138].
3.4
Maximum-Likelihood-Based Methods
Maximum Likelihood (ML)-based approaches were first advocated and popularized
by R. A. Fisher around 1920, but were used in estimation of evolutionary trees for
the first time by Edwards and Cavalli-Sforza [38]. Joseph Felsenstein published an
algorithm for computing the likelihood P (D|T ) for discrete and continuous characters in 1973 [43, 44], where D is the gene sequence information, and T is the
topology and branch-lengths of the gene tree. Given the topology of gene tree and
the branch lengths, the likelihood of the sequences is computed starting from the
leaves by marginalizing all possible character assignments of the internal vertices.
Finding the optimal ML tree, however, still remains a problem as it is NP-complete
[47]. ML-based methods are found to be more accurate in majority of cases in comparison with distance-based methods [47]. Getting the ML-tree usually involves
many iterations of tree and branch lengths optimization. The tree optimization is
done by methods such as branch swapping (e.g., nearest neighbor interchange and
subtree pruning and re-grafting). One may also use bootstrapping to estimate the
confidence intervals of the inferred parameters. Some of the popular programs that
implement ML-based methods include RAxML [132], PhyML [58], PAML [156],
and PLL [50].
26
3.5
CHAPTER 3. MODELING EVOLUTION
Bayesian Methods
Bayesian inference methods are based on Bayes’ rule which states that
P r(θ|D) =
P r(D|θ)P r(θ)
P r(D)
where P r(θ|D) is the posterior distribution about the set of hypotheses, P r(θ)
is the prior belief about the parameters θ, P r(D|θ) is the likelihood of the data
D given the parameters θ, and P r(D) is the probability of data (which acts as
a normalization constant). The posterior distribution is directly proportional to
the product of the likelihood of the data and the prior. When the normalization
constant (P r(D)) involves high-dimensional integrals over all possible values of parameters θ, the posterior distribution is hard to compute directly. In order to obtain
posterior distribution using Bayesian methods, phylogenetic methods usually relies
on MCMC-based approaches (MCMC-based methods are discussed later in this
section). The use of priors is, however, controversial and has been debated in phylogenetics. The subjective nature of prior makes the Bayesian inference subjective.
This is, however, also an advantage of Bayesian inference as it allows the researcher
to incorporate prior biological knowledge and underlying assumptions in the form of
an appropriate prior. It may be difficult, due to practical reasons, to provide appropriate prior in terms of a probability distribution representing existing information
about the data. It has been shown that under certain conditions more and more
similar posterior distributions are obtained with increasing amount of data [12],
therefore, priors do not always have strong influence on results. Non-informative
priors are argued as an objective choice, when there is lack of knowledge about the
prior. The posterior density may still be sensitive to the boundary conditions.
Markov Chain Monte Carlo (MCMC)
Markov Chain Monte Carlo (MCMC)-based inference in phylogenetics is used to
find the relative probabilities of the possible evolutionary scenarios. MCMC is typically used in cases where one has to search high dimensional spaces and analytical
solutions are not easily available. This method sets up a Markov Chain, where every state are points in the parameter space and only depends on the previous state.
The parameter space is sampled by a random walk in which the frequency of state
visits, in the limit, equals the posterior distribution. The posterior distribution is
then summarized to get the estimates of the marginal distribution for the quantities of interest. Metropolis-Hastings is a sampling method used in cases where
direct sampling from posterior distribution is hard. In every iteration, a new state
x0 is proposed by perturbing one or many parameters of the current state x. The
new state is then accepted with a probability that is determined by comparing the
likelihoods of the current state and the proposed state with respect to the desired
distribution assuming the forward/backward proposal probabilities between any
3.6. RECONCILIATION-BASED APPROACHES
27
two states are the same. If the forward/backward proposal probabilities between
the two states are not the same, then the acceptance probability of any proposed
state is given by:
)
(
T (x0 , x) P (x0 )
0
A(x, x ) = 1,
T (x, x0 ) P (x)
where T (x, x0 ) is the transition probability given by the proposal function from state
x to state x0 . Although the random walk process eventually converges to the desired stationary distribution, the initial samples may follow a different distribution.
These initial samples, also known as the burnin period, are therefore conventionally
discarded. As more and more states are sampled, the distribution of sampled states
becomes an increasingly better approximation of the desired distribution. This
method has theoretical guaranties on convergence to the posterior distribution and
is used frequently in phylogenetics, as well as in other fields. MCMC-based methods are computationally demanding but the availability of improved computational
power and parallelization are making them increasingly feasible. In the particular
case of gene tree reconstruction, proposal functions such as branch swapping are
used to sample the search space. Some Bayesian phylogenetic tools that employ
MCMC framework to reconstruct the gene trees are MrBayes [72], PrIME [1], BAliPhy [135], PhyloBayes [89], BEAST [36], and JPrIME [129]. In this thesis, I will
discuss some novel methods that use the MCMC framework to reconstruct gene
trees and reconcile those gene trees with the corresponding species tree, which enables us to get the precise time estimates of the events (e.g., duplications, losses,
LGT or speciations) represented by gene tree vertices.
3.6
Reconciliation-based Approaches
Reconciliation-based approaches maps the evolutionary events of a gene tree to
that of the species tree. Depending on the model, evolutionary events of a gene
tree are typically duplications, losses, or lateral gene transfers. As discussed earlier,
the evolutionary history of a gene family is represented by a gene tree, while its
internal vertices represents the evolutionary events. The internal vertices of the gene
tree are mapped to the edges or vertices of the species tree, thereby determining
the evolutionary events. The reconciliation-based approaches expects the species
phylogeny and the mapping of extant genes to the leaves of species tree as input
in addition to the MSA of the gene family, and are treated as reality. Two genes
are orthologous if their least common ancestor happens to be due to a speciation
event. In all other cases, where the least common ancestor happens on an edge of
the species tree, the two gene lineages are paralogous to each other. An internal
vertex of the gene tree may also be a transfer event, in which case it is usually
mapped to any two edges of the species tree. Similarly, an internal vertex may also
be a pseudogenization event in the gene tree. Pseudogenization vertices are degreetwo vertices and have only one descendant lineage, representing pseudogenization
of the gene lineage (See Paper II).
28
CHAPTER 3. MODELING EVOLUTION
Reconciling the evolutionary history of genes with that of species is not a trivial
task. Gene-species tree discordance can arise, e.g., because that genes usually do
not have one to one correspondence with the species. The process of reconciliation
is complicated by evolutionary events such as gene duplications and gene losses.
Moreover, the lateral gene transfers is another phenomenon that adds to the genespecies tree discordance specially in the prokaryotes. Some other such processes
are hybridization of species, gene fusion/fission that makes the process of reconciliation even harder. Hybridization of species is common in plants, but also found in
animals, which sometimes can result in multiple sets of homologous chromosomes.
last decade
(see, for instance, [138–142]). In parsimony, LGT adds a level of
In gene fusion two gene lineages merge into one, which cannot be appropriately
complexity
compared
to structures.
gene duplication
and on
loss,
because
multiple
modeled
by tree-like
Gene fission,
thepartly
other hand,
splits
a gene MPR
two coexist.
or more parts.
Incomplete
sorting such
(ILS) as
also
makesapplicable
the processonly on
optimasinto
often
Initially
having lineage
limitations
being
of reconciliation further complicated. ILS happens when a species lineage, having
monocopy
families
allowing
time,speciation
there are
today
population
withor
more
than oneLGT
alleles,backwards
goes throughinseveral
events,
andefficient
than
one alleles
survives in loss
the descendent
lineages. sound
In such LGT
cases, the
allele for inmethodsmore
that
model
duplication,
and temporally
events,
tree might be different from the gene tree, and therefore has the potential to cause
stance the
work by Tofigh and Lagergren [143], David and Alm [144], Bansal et
discordance.
al. [145,146], Daubin et al. [28,147], and Doyon et al. [148,149]. A major shortparsimony has been
used for reconciling
the evolutionary
history
of reconcomingTraditionally,
of many combinatorial
frameworks
is that sequence
evolution
and
genes with that of species. Methods based on the parsimony principle are quick and
ciliationefficient
constraints
not taken
into scenarios
account[45,
simultaneously.
LGT from
but can are
be misleading
in some
95]. Probabilistic Also,
approaches
aimed
at
reconciling
possible
gene
trees
with
the
given
species
tree
have
been
species not included in the analysis can complicate gene family analysis,ineither by
troduced in the last decade. These approaches are biologically more realistic, and
introducing
homologs alongside those stemming from the most recent common
mathematically robust. They are typically Bayesian-based and as discussed earlier
ancestorthey
of provide
the studied
species,
or by giving
to the ofentire
family.
a posterior
distribution
over therise
parameters
the model.
I will briefly
discuss both of the approaches in the following:
Figure 3.1: Two ways to reconcile a gene tree with species tree [128]
Figure 10. Shows two ways to reconcile the same gene tree topology. The non-MPR
solution induces two implicit losses (note that most methods work with the pruned
topology only).
Not surprisingly, there is a twilight zone between purely combinatorial and probabilistic approaches. For example, sampling among equally optimal solutions in
LGT-enabled parsimony to achieve better estimates was suggested recently [146].
Another approach is to let duplication-loss reconciliations guide improvements of
an a priori estimated tree inferred using sequence data [150–152].
3.6. RECONCILIATION-BASED APPROACHES
29
Parsimony-Based Approaches
Goodman et al [56] pioneered the field by introducing the concept of reconciliation.
They used a parsimony-based approach and proposed an algorithm that finds the
most parsimonious reconciliation (MPR). The MPR is a reconciliation that uniquely
maps the vertices of a gene tree to the vertices or edges of a species tree such that
the number of inferred evolutionary events (such as duplications and losses) is
minimized. Many variations of this algorithm has been proposed since then. In
figure 3.1, the reconciliation on the left minimizes the number of evolutionary
events and is therefore, MPR. The reconciliation on the right have two additional
gene losses, a gene duplication, and speciation event (six events in total) while the
reconciliation on the left has only two evolutionary events. MPR works under the
assumption that evolutionary events are rare and therefore, parsimonious scenarios
are the most likely scenarios. The MPR-based methods are faster but are less
realistic than probabilistic methods.
A number of parsimony-based methods have been introduced that reconciles the
gene tree with the species tree. These methods take into account one or more of
evolutionary events such as gene duplications, gene losses, lateral gene transfers,
and hybridization of species. In the following I will briefly discuss some recent
parsimony-based methods that takes into account 1) duplication and loss events
only, 2) lateral gene transfers, 3) incomplete lineage sorting and 4) hybridization
events.
Many parsimony-based reconciliation methods, proposed in the past decade, take
into account gene duplications and gene losses while reconciling a given gene tree
with a given species tree. One such approach was proposed by Durand et al. [37].
This method reconstructs the gene tree by any sequence evolution model of choice
in the first stage. In the second stage, the gene tree is refined by rearranging
regions of the tree that do not have strong support, in order to minimize the cost
of duplication and losses. Some commonly used parsimony-based reconciliation
methods are SYNERGY [146] and NOTUNG [145].
Hallet et al. [63, 144] proposed one of the first parsimony methods that take into
account lateral gene transfers. They proposed fixed parameter tractable algorithms
that count the minimum number of gene duplications and lateral gene transfer
events required by any reconciliation. Recently, Doyon et al. [34] proposed a
combinatorial parsimony-based approach (Mowgli) for reconciling a gene tree and
a species tree that computes a time-consistent (i.e. lateral gene transfers can happen
in the same time slice) most parsimonious reconciliation. Other methods addressing
lateral gene transfers along with gene duplications and gene losses as evolutionary
events include AnGST [27] and RANGER-DTL [9].
Incomplete Lineage Sorting is another evolutionary scenario that contributes to the
discordance between gene trees and species trees. Wu et al. [155] recently proposed
30
CHAPTER 3. MODELING EVOLUTION
an algorithm (DLCpar) to incorporate deep coalescence that uses an additional
allele tree to find the most parsimonious reconciliation in the presence of coalescent
and duplication-loss history. Each duplication creates an additional locus that is
not linked with the original locus. DLCpar uses a data-structure of three nested
trees, i.e., the allele tree evolving inside the gene tree (called locus tree by Wu et
al.), which evolves inside the species tree, together with a dynamic programming
approach, and efficiently explores the entire search space of reconciliations, and
provide the most parsimonious reconciliation as output.
Hybridization of species is yet another factor that contributes to the discordance
between gene/allele tree and species tree. A parsimony-based approach for reconciling gene trees within the branches of a phylogenetic network was recently proposed
by Yu et al. [157], where they considered speciation, hybridization and incomplete
lineage sorting as the evolutionary events affecting the gene tree evolution. The
species tree is first converted into a multi-labeled tree (introduced in [71]), which
makes it easier to apply all existing tree-based phylogenetic methods in the context of hybridization. A phylogenetic network is converted into a multi-labeled tree
by going in a bottom-up fashion and replicating the subtrees at all the reticulation nodes. This results in a rooted binary tree, whose leaves are not necessarily
uniquely labeled.
Probabilistic Approaches
A number of generative models have been proposed for the evolution of genes inside
a species tree. The Duplication-Loss model is one such model, which models the
evolution of gene tree inside the species tree. Gene lineages are duplicated and
lost during the evolution according to an underlying birth-death process. The
Duplication-Loss model has been frequently used to model evolutionary events (e.g.,
speciation, duplications, LGT) in trees. The gene duplications and gene losses
were first modeled probabilistically by a Duplication-Loss model. DuplicationLoss models typically employ a birth-death process that models the genes evolving
inside a species tree. The generalized birth-death process was introduced by David
G. Kendall [85]. A birth-death process is initiated at the root of the species tree
that undergoes births (gene duplications) and deaths (gene losses). A birth in a
lineage cause duplication, which bifurcates a gene lineage. A loss of a gene lineage
is modeled by a death in the birth-death process. Gene lineages passing through
speciation vertices in the species tree always bifurcate. At the end of the birth-death
process, all the gene lineages having no descendants in extant species are pruned.
Duplication-loss models resolve the discordance caused due to duplications and
losses of genes in the evolutionary history, and do not consider the population level
processes that cause discordance, such as incomplete lineage sorting. Some recent
methods using Duplication-Loss models in different settings will be discussed later
in this section.
3.6. RECONCILIATION-BASED APPROACHES
31
Coalescent models are another type of generative models, that are used for tracing
back the genealogical history of alleles. These models are retrospective models for
population genetics that model the evolution of alleles backwards in time. For a
given locus in the genome, at any given time there may be several alleles due to
mutations in some of the gene copies across individuals of the population. The
evolutionary history of the alleles can be represented as an allele tree. Once the
speciation happens, the alleles are sorted randomly, and as a result the resulting
populations of species receives all or a subset of the parental species’ alleles. Successive speciation events over a brief interval of time may result in the difference in
the topology between the alleles tree and the species tree [120]. Use of coalescent
models for resolving this problem has became the method of choice. Coalescent
models builds on the Wright-Fisher model, which was independently developed by
Sewall Wright and Ronald Fisher during 1929-31. This model considers a population that is equally likely to mate with each other, with the same reproductive
success, and evolving in the same conditions. The alleles may be haploid - having
one parent, or diploid - having two parents. The population of N genes evolves in
discrete generations. The process runs backwards in time, and each gene randomly
chooses its parent from the parental generation independently from other genes
in the same generation. An alternative model for the problem was introduced by
Patrick Moran in 1958. The generations of alleles are allowed to overlap in this
model, and population size remains constant. At discrete time intervals, two individuals of population are randomly chosen for reproduction and death. Some
recent methods using the coalescent models will be discussed later in this section.
In the following I will discuss some recent probabilistic models that are used to
resolve gene-species inconsistencies by taking into account evolutionary events such
as gene duplications, gene losses, incomplete lineage sorting and hybridization.
Duplication-Loss Models
The first probabilistic model that modeled duplications and losses was proposed
by Arvestad et al. in 2003 [3], and modeled how a gene family evolves inside
a species tree by undergoing evolutionary events such as gene duplications and
gene losses. In 2004, the same group introduced a gene sequence evolution model,
that integrated probabilistic model (DLRS) for gene duplications, gene losses and
sequence evolution under a molecular clock [4]. The assumption of molecular clock
was later relaxed [1]. Figure 3.2 illustrates the basic components of the model, a
more detiled description are given in paper I. The approach takes an alignment of
gene sequences, a dated species tree, and a mapping of the genes to the species
as input. A Bayesian method (MCMC-based framework) is then used to obtain
a posterior distribution over all gene trees, edge lengths over the gene trees, and
32
CHAPTER 3. MODELING EVOLUTION
Figure 3.2: An illustration of DLRS Model, where A) illustrates the evolution of a gene
lineage modeled according to the Duplication-Loss model that undergoes duplications and
losses. B). Illustrates the clock relaxation according to the Rate model, C) Illustrates the
Sequence evolution model
other parameters. The probability density in this case is given as follows:
p(G, l, θ|D, S) =
P (D|G, l)p(G, l|θ, S)p(θ)
P (D|S)
whereas G is the gene tree, l are edge lengths over gene tree, D is the multiple
sequence alignment of gene sequences, S is the dated species tree, and θ denotes
the parameters of Duplication-Loss model and substitution rate model. The first
factor P (D|G, l) can be computed using the standard DP algorithm by Felsenstein
[46]. The second factor p(G, l|θ, S) is the probability of a gene tree (along with
the branch-lengths) evolving inside a species tree given all the parameters of the
model. This factor can be computed using the dynamic programming algorithm
introduced in [1]. The priors p(θ) are chosen so that they can be easily computed.
The denominator P (D|S) cancels during the estimation of posterior probabilities
in in any ratio between two such probabilities.
Instead of searching the entire search space, some methods restrict the search space
to near-MPR gene trees. This reduces the required computation and accuracy
appears not to be compromised in a majority of the cases [117, 116]. Both of the
above-mentioned methods assume that the correct species tree is known. Recently,
Boussau et al. [15] proposed a probabilistic method that estimates the species tree
and gene tree at the same time. The input to this method are a multiple sequence
3.6. RECONCILIATION-BASED APPROACHES
33
alignment of gene sequences, species names and a mapping between genes and
species, while the method infers gene trees, species tree and branch-wise numbers
of duplications and losses. The following equation is maximized for gene tree G
and species tree S:
Y
L(Gi )
L(G, S, N |A) =
Gi ∈G
whereas G is the set of all gene families, and the N is the branch-wise expected
numbers of duplications and losses. The likelihood for a gene family is given as:
L(Gi ) = L(S, N |Gi )L(Gi |Ai )
where Gi and Ai are the gene tree and alignment for a gene family G, respectively.
The likelihood L(Gi |Ai ) can be computed using any probabilistic model of sequence
evolution [47]. The factor L(S, N |Gi ) is the likelihood of the reconciliation of a gene
tree with a species tree according to duplication loss rates. This factor is computed
in a similar way to Åkerborg et al. [1], but the gene families share the parameters of
duplication and loss events that are branch-specific on the species tree. The method
does not take into account the branch-lengths on the gene trees while reconciling
a gene tree with a species tree, and instead only reconciles the topologies. This
reduces the number of global parameters and therefore helps in faster computation.
But since the hierarchical model also includes sequence evolution for joint inference
of gene trees and species trees, branch lengths are essential to estimate.
LGT Models
Horizontal gene transfer or lateral gene transfer is another evolutionary event that
can create discordance between gene trees and species trees. Sjöstrand et al.
[143, 128] introduced one of the first probabilistic models, DLTRS (Duplication,
Loss, Transfer, substitution Rate variation over gene tree, and Sequence evolution),
that integrates duplication-loss events, lateral gene transfer events along with sequence evolution in a single comprehensive model. A modified version of the birthdeath process is used in this case, which models lateral gene transfers as well as
gene duplications and gene losses. The probability of a gene tree, its edge lengths
and other parameters are computed similarly to what is done in Åkerborg et al. [1].
In this case, the gene tree lineages are allowed to jump across the species tree lineages in the same time (see Figure 3.3). The present work extends this method and
propose a dynamic programming algorithm for sampling reconciliations and computing most-likely reconciliations. Further details about this method are given in
chapter 4, and in Paper III. Another method was proposed by Suchard [134], where
he proposed a hierarchy of stochastic models to investigate LGTs using multiple
orthologous gene alignments. The models works in a hierarchical manner, in which
the top layer builds a joint distribution over multiple gene trees and an unknown
species tree through a random walk in the ‘trees space’. In the bottom layer, gene
34
CHAPTER 3. MODELING EVOLUTION
Figure 3.3: PrIME-DLTRS [130]
trees are reconstructed given the multiple sequence alignments conditional on the
random walk process. This hierarchical distribution provides likelihoods of gene
trees given an unknown species tree and unknown number of LGTs. Szollosi et al.
[139] integrated the processes of originations (from ‘scratch’ or from extinct species),
duplications, losses and lateral gene transfers into single model, ODT (Origination,
Duplication, Transfer, and Loss) to reconstruct a chronologically ordered species
tree by explicitly modeling the evolution of genes in their genomes. The birthdeath process in this case consists of two types of births, i.e. duplication, or lateral
gene transfer. Moreover, in this model lateral gene transfer can also occur from
the species that have gone extinct (or not present in the species under consideration). The probability of a gene tree G given a species tree S, with probabilities of
origination, duplication, transfer and losses denoted by M is computed as follows:
!
X
X
~
p(G|S, M) =
PO (x)P (R, x)
~
R∈R
x∈N (S)
where R is the set of all possible roots of G, N (S) denotes all positions across
species tree, and pO (x) corresponds to the origination probability at position x
along S. Under the assumption that the gene families G evolved independently,
the likelihood of S and M is computed as:
Y
LODT (S, M|G) =
p(G|S, M)
G∈G
In the context of prokaryotes, models that take into account lateral gene transfer
along with gene duplications and losses are biologically more realistic than the
Duplication-Loss models. However, the addition of lateral gene transfer to the
Duplication-Loss models usually comes at the cost of increase in the amount of
required computation.
ILS Models
The probabilistic methods discussed above do not account for Incomplete Lineage
Sorting (ILS) of alleles. Different coalescent models have been developed that
3.6. RECONCILIATION-BASED APPROACHES
35
account for variation of population sizes, divergence times, and migration rates
[91, 115, 69]. Incomplete lineage sorting has recently been addressed by using
Figure 3.4: Duplication and loss events within a multispecies coalescent. (A) A duplication
occurs in one chromosome and creates a new locus, “locus 2”, in the genome. At locus
2, the Wright-Fisher model dictates how the frequency p of the child duplicate (black
dots) competes with the null allele (white dots) until it eventually fixes (p = 1). A
allele tree is therefore a “traceback” in this combined process. (B) A new duplicate can
undergo hemiplasy, and fixes in some lineages and goes extinct in others. (C ) Similar
to duplication, a gene loss (deletion) starts in one chromosome and drifts until it fixes or
goes extinct. [118]
coalescent models along with Duplication-Loss models. Rasmussen et al. [118]
proposed a probabilistic model DLCoalRecon, that integrates the Duplication-Loss
model with coalescence model. DLCoalRecon expects a dated species tree S, an
estimated population size N , gene duplication-loss rates (λ and µ), and allele tree
topology T A as input (Note that I use different terminology for allele tree and gene
tree than [118]). The gene tree T G is generated by a birth-death process similar
to [1]. For each duplication node in the gene tree, one of the children is randomly
denoted as a child and the other as the parent. Each new gene (child) corresponds
to a new chromosomal location. The allele tree T A is generated bottom-up using
a multilocus coalescent within the gene tree T G (see Figure 3.4). A hill-climbing
search is used to estimate the maximum a posteriori reconciliation between the
three trees. It returns as output the reconciliation R, which is defined as tuple,
R = (T G , RA , RG , δ G ), where RA is a mapping of the allele tree to the gene tree,
RG is a mapping from the gene tree to the species tree S, and δ G is the set of
child nodes. The probability of a gene tree T G and the reconciliation R given the
dated species tree (S, tS ), population sizes and and duplication-loss rates is given
as follows:
P (T A , R) = P (δ G |T G , RG , S)P (T G , RG |S, θ)×
Z
P (T A , RA |T G , tG , σ G , N G )P (tG |T G , RG , S, θ)dtG
G
G
(3.1)
The first factor P (δ G |T G , RG , S) = 2−|dup(T ,R ,S)| where dup(T G , RG , S) gives
the number of duplications in the gene tree. Second factor, P (T G , RG |S, θ) is com-
36
CHAPTER 3. MODELING EVOLUTION
puted similar to [1], and the integration of the last factor is the main contribution
of this method.
Hybridization Models
As discussed earlier, hybridization of species may result in a non-tree like structure
or cause discordance between the gene/allele tree and the species tree [125, 14].
Several methods have attempted to resolve the discordance/non-tree like evolution
caused by hybridization by using a species network, where an allele in the hybrid
species is assumed to originate from one of the two parental species with some
probability. Meng and Kubatko [100] in 2009 proposed a method that estimates
the role of hybridization using a likelihood as well as bayesian framework, both
in case of hybridization and coalescence. The model assumes that in the case of
hybridization of species, a randomly selected gene from one of the two parental
species is inherited. A usual coalescence process is employed for the alleles, where a
parent for hybridized allele is chosen according to the probability γ, which is inferred
based on a likelihood function. Yu et al. [158] recently proposed another approach
for addressing hybridization of species in the presence of incomplete lineage sorting.
The phylogenetic network W is first converted into a multi-labeled species tree T ,
(a tree whose leaves are not uniquely labelled by a set of taxa) by going in a bottomup fashion and replicating the subtrees at all the reticulation nodes. This results in
a rooted binary tree, whose leaves are not necessarily uniquely labeled. Alleles of
the tips of gene tree G are mapped in every valid way to the tips of T , and finally
the probability of G is computed as the sum, over all valid allele mappings, of the
probabilities of G given T .
P (G|W, l, γ) =
X
P (G|T, l0 , γ 0 , f )
f ∈F
where edge lengths of the gene tree are denoted by the vector l, of hybridization
probabilities (which indicates for each allele in a hybrid population its probability of
inheritance from each of the two parent populations) are denoted by γ, and all valid
allele mappings are denoted by F. Given a collection (G) of gene tree topologies,
one per locus, in a set of sampled loci, where pg is the sum of posterior probabilities
associated with the gene tree g, the likelihood function is given by:
L(W, l, γ|G) =
Y
[P (G = g|W, l, γ)]pg
g∈G
Like other coalescent models, population size is assumed constant and it is assumed
that no recombination takes place.
3.6. RECONCILIATION-BASED APPROACHES
37
Pseudogenization Models
Gene losses is also a factor that contributes towards the discordance in the reconciliation of gene trees and the species tree. The process of pseudogenization has
been regarded as the most likely fate for the newly duplicated genes. Interestingly, some pseudogenes remain conserved over long period of time and are even
attributed a potential function. The conservation of pseudogenes over long period
of time and their potential function highlights the importance of the studying pseudogenes. Svensson et al. [137] conducted a genome-wide survey for and analysis
of human pseudogenes, in which they used a BLAST-based pipeline augmented by
maximum-likelihood phylogeny estimations. Their results suggested that functional
pseudogenes are rarely found in the evolutionary timescale, but they still exists. A
comprehensive model that models pseudogenization events, gene duplications, gene
losses, and sequence evolution under a relaxed molecular clock is introduced in paper II. The model is based on a birth-death process, where a gene family starts with
a gene at the root, and may pseudogenized during evolution. A pseudogene lineage
may duplicate or get lost just like a gene lineage. Conversion of a pseudogene to
gene is however not allowed in the model. A neutral codon substitution model is
used for the sequence evolution of pseudogenes. Paper II provides the details of
this model.
Chapter 4
Present Investigations
Paper I: In this study, we extend the DLRS framework introduced by Åkerborg
et al. [1]. The proposed method samples reconciliations as well as compute
the most likely reconciliation from the posterior distribution over gene trees
and other parameters. The method is used to analyze genome-wide gene
families of OPTIC dataset consisting of nine vertebrate species. The species
tree is augmented by a heatmap for each edge, illustrating how frequently
duplications occur on the edges, across the gene families. We also compare the
obtained reconciliations with the most parsimonious reconciliations (MPR)
and conclude that MPR leads to an incorrect reconciliation in 19% of all
reconciliations. Finally, we propose algorithms for sampling and computing
the most likely realizations (a finer reconciliation, that maps vertices of the
gene tree to specific time points on the species tree).
Paper II: In order to facilitate the proper analysis of gene families including the
process of pseudogenization, we introduce PDLRS by introducing the possibility of pseudogenization of gene lineages. The model integrates sequence
evolution (under a relaxed molecular clock for substitution rates) with evolution of genes and pseudogenes that may undergo duplication/loss events,
gene-to-pseudogene conversions according to a birth-death process. We devise MCMC-based methods that allows data analysis with respect to this
model, and apply it on synthetic as well as biological datasets. The biological
datasets consists of genes and pseudogenes from two largest gene families in
vertebrates, i.e. Olfactory Receptors and Zinc Fingers. Olfactory receptors
are studied across human, dog, opossum, and platypus, while zinc fingers are
studied across the four primate species human, chimpanzee, rhesus monkey,
and orangutan.
Paper III: This study extends the model introduced by Sjöstrand et al. [128].
The proposed method provides algorithms for sampling reconciliations as well
39
40
CHAPTER 4. PRESENT INVESTIGATIONS
as computing most likely reconciliations from the posterior distribution over
gene trees, edge lengths and other parameters of the model. By conducting
a simulation study, we observe that the method is able to infer the true LGT
events on gene tree and reconcile it to the correct edges of the species tree in
most cases.
Chapter 5
Discussion & Conclusion
The increasing number of fully sequenced genomes have expanded our understanding about the evolutionary process. Identifying the regulatory, functional and nonfunctional regions of the genomes is an important step in understanding genomes.
This valuable insight is provided by comparing genomes or large part of genomes
across the species. Phylogenomics is the field in which we compare genomes to
reconstruct their evolutionary history. It also helps us understand when and how
different genes evolved inside different species. This enable us to study the origins of
genetic diseases that are incurable at the moment. Furthermore, phylogenomics also
open up new avenues to understand diseases like cancer, alzheimer etc. and other
natural processes at the level of our genome such as aging,
by comparing genomes across different species.
In this thesis, two methods are proposed to estimate the age of evolutionary events. In the
first project, age of the gene duplications are estimated on evolutionary scale, while in the last
project we estimate age of the gene duplication
events as well as gene transfer events. Mapping
the gene duplication events to the evolutionary
time-scale give us interesting insights into the
evolutionary history and the ecology of earth.
Heatmap of the gene duplications across genefamilies for the selected species of vertebrates is
given in figure 5.1. The heatmap provides an
overall picture of the rate of gene duplications
across different lineages. We observed the high- Figure 5.1: Heatmap of gene dupliest rate of duplication during the Cambrian pe- cation events across gene families
riod, followed by the Ordovician period. Most of
the major animal phyla resembling current day
life forms appeared during the Cambrian period.
41
42
CHAPTER 5. DISCUSSION & CONCLUSION
The ordovician period, which is associated with a mass extinction, also had a high
rate of gene duplication according to our analysis. High duplication rates were also
observed for the edges corresponding to the Carboniferous period, which is associated with highest levels of oxygen on earth, and Permian period, which is associated
with mass extinction and plummeting oxygen levels [73]. Moreover, a high rate of
duplication was also observed for the common ancestral edge of human, mouse and
dog (Boreoeutheria). This edge corresponds to the Cretaceous period which is also
associated with a mass extinction [75]. The role of extinctions in shaping evolution
has been under debate [119, 76]. It would be interesting to investigate further if
a higher rate of duplication on certain vertebrate lineages is related to the mass
extinctions or not. Comprehensive probabilistic methods that model most of the
evolutionary events can give us further insight into the evolutionary history of the
species and their ecology.
Modeling pseudogenization events was the second project of this thesis. Pseudogenes were initially thought to be the ‘junk DNA’ but recently many studies have
reported their potential functionality. Some have even called them ‘potogenes’ suggestive of the potential functionality they might have [7]. A model that can take
into account the process of pseudogenization realistically can be very helpful in correctly determining the evolutionary history of pseudogenes. Pseudogenes that have
remained conserved across large evolutionary distance are more likely to be functionally active. Identifying such pseudogenes is obviously important to understand
the functioning of genomes. In this project we introduce a probabilistic method
that estimates the age of the gene-to-pseudogene conversion points along with the
gene duplication events on the evolutionary scale. Two gene families were studied using this method and some ancient pseudogenes in the selected subfamilies of
olfactory receptors and zinc-finger-gene superfamilies were identified. Further analysis of more gene families can be helpful to reconstruct the evolutionary history of
genes and pseudogenes more precisely. Taking into account the ecology of species,
it can be further investigated how much the genomes of different species have been
shaped by the ecology.
The methods presented in this thesis have the room for improvement, as they do
not take into account all the evolutionary processes. Evolutionary processes such as
incomplete lineage sorting and hybridization are two important evolutionary processes that plays an important role in the evolution of a genome. Population-based
phylogenomics will become more and more feasible with the increasing number
of sequenced genomes per species. Hybridization of genes is also an important
confounding factor while reconstructing the gene phylogenies. There are also evolutionary processes acting at sub-gene level. Some genes are composed of multiple
protein domains. The evolutionary processes on the domain level, might also play
an important role in the evolution of a gene family. Having a comprehensive model
that takes into account most of the evolutionary mechanisms is therefore, an obvious
future direction. However, the models that take into account more and more evo-
43
lutionary processes have the possibility to suffer from over-parameterization. This
makes the computational aspect of this interesting problem even more challenging.
With the availability of computational resources and sequenced genomes with better coverage, it is relatively easier to do the phylogenomic analysis. However, at
the same time there is a need for comprehensive probabilistic methods that can efficiently use the genomic information to reconstruct the evolutionary history. This
thesis adds new methods to estimate the approximate times for the evolutionary
events to the existing existing literature. Age of the gene duplication, lateral gene
transfer as well as gene-to-pseudogene conversion can now be estimated as well
using these methods. Combining genome-wide phylogenomic studies with the geological history of earth constitute a promising means to increase our understanding
about the process of evolution.
Bibliography
[1]
Örjan Åkerborg, Bengt Sennblad, Lars Arvestad, and Jens Lagergren. Simultaneous Bayesian gene tree reconstruction and reconciliation analysis. Proceedings of the National Academy of Sciences, 106(14):5714–5719, 2009.
[2]
Stephen F Altschul, Warren Gish, Webb Miller, Eugene W Myers, and
David J Lipman. Basic local alignment search tool. Journal of molecular
biology, 215(3):403–410, 1990.
[3]
Lars Arvestad, Ann-Charlotte Berglund, Jens Lagergren, and Bengt
Sennblad. Bayesian gene/species tree reconciliation and orthology analysis
using MCMC. Bioinformatics, 19(suppl 1):i7–i15, 2003.
[4]
Lars Arvestad, Ann-Charlotte Berglund, Jens Lagergren, and Bengt
Sennblad. Gene tree reconstruction and orthology analysis based on an integrated model for duplications and sequence evolution. In Proceedings of the
eighth annual international conference on Resaerch in computational molecular biology, pages 326–335. ACM, 2004.
[5]
Oswald T Avery, Colin M MacLeod, and Maclyn McCarty. Studies on the
chemical nature of the substance inducing transformation of pneumococcal
types induction of transformation by a desoxyribonucleic acid fraction isolated
from pneumococcus type III. The Journal of experimental medicine, 79(2):
137–158, 1944.
[6]
Alexander V Badyaev. Stress-induced variation in evolution: from behavioural plasticity to genetic assimilation. Proceedings of the Royal Society
B: Biological Sciences, 272(1566):877–886, 2005.
[7]
Evgeniy S Balakirev and Francisco J Ayala. Pseudogenes: are they "junk" or
functional DNA? Annual review of genetics, 37(1):123–151, 2003.
[8]
Evgeniy S Balakirev, VR Chechetkin, VV Lobzin, and Francisco J Ayala.
DNA polymorphism in the β-esterase gene cluster of Drosophila melanogaster.
Genetics, 164(2):533–544, 2003.
45
46
BIBLIOGRAPHY
[9]
Mukul S Bansal, Eric J Alm, and Manolis Kellis. Efficient algorithms for the
reconciliation problem with gene duplication, horizontal transfer and loss.
Bioinformatics, 28(12):i283–i291, 2012.
[10]
David Benovoy and Guy Drouin. Processed pseudogenes, processed genes,
and spontaneous mutations in the Arabidopsis genome. Journal of molecular
evolution, 62(5):511–522, 2006.
[11]
Joseph P Bielawski and Ziheng Yang. Maximum likelihood methods for detecting adaptive evolution after gene duplication. In Genome Evolution, pages
201–212. Springer, 2003.
[12]
David Blackwell and Lester Dubins. Merging of opinions with increasing
information. The Annals of Mathematical Statistics, pages 882–886, 1962.
[13]
Christopher Blair and Robert W Murphy. Recent trends in molecular phylogenetic analysis: where to next? Journal of Heredity, page esq092, 2010.
[14]
Erik W Bloomquist and Marc A Suchard. Unifying vertical and nonvertical
evolution: a stochastic arg-based framework. Systematic biology, 59(1):27–41,
2010.
[15]
Bastien Boussau, Gergely J Szöllősi, Laurent Duret, Manolo Gouy, Eric Tannier, and Vincent Daubin. Genome-scale coestimation of species and gene
trees. Genome research, 23(2):323–330, 2013.
[16]
MG Bulmer. Francis galton. Pioneer of Heredity and Biometry. The, 2003.
[17]
Ze Cheng, Mario Ventura, Xinwei She, Philipp Khaitovich, Tina Graves,
Kazutoyo Osoegawa, Deanna Church, Pieter DeJong, Richard K Wilson,
Svante Pääbo, et al. A genome-wide comparison of recent chimpanzee and
human segmental duplications. Nature, 437(7055):88–93, 2005.
[18]
Ramu Chenna, Hideaki Sugawara, Tadashi Koike, Rodrigo Lopez, Toby J
Gibson, Desmond G Higgins, and Julie D Thompson. Multiple sequence
alignment with the clustal series of programs. Nucleic acids research, 31(13):
3497–3500, 2003.
[19]
Francis S Collins, Michael Morgan, and Aristides Patrinos. The human
genome project: lessons from large-scale biology. Science, 300(5617):286–290,
2003.
[20]
1000 Genomes Project Consortium et al. An integrated map of genetic variation from 1,092 human genomes. Nature, 491(7422):56–65, 2012.
[21]
ENCODE Project Consortium et al. An integrated encyclopedia of DNA
elements in the human genome. Nature, 489(7414):57–74, 2012.
BIBLIOGRAPHY
47
[22]
Diego Cortez, Patrick Forterre, Simonetta Gribaldo, et al. A hidden reservoir
of integrative elements is the major source of recently acquired foreign genes
and ORFans in archaeal and bacterial genomes. Genome Biology, 10(6):R65,
2009.
[23]
Francis Crick. Central dogma of molecular biology. Nature, 227(5258):561–
563, 1970.
[24]
Francis H Crick. On protein synthesis. In Symposia of the Society for Experimental Biology, volume 12, page 138, 1958.
[25]
Charles Darwin. On the origin of species london. UK: John Murray, 1859.
[26]
Charles Darwin. On the origins of species by means of natural selection.
London: Murray, 1859.
[27]
Lawrence A David and Eric J Alm. Rapid evolutionary innovation during an
archaean genetic expansion. Nature, 469(7328):93–96, 2011.
[28]
Margaret O Dayhoff and Robert M Schwartz. A model of evolutionary change
in proteins. In In Atlas of protein sequence and structure. Citeseer, 1978.
[29]
Hugo De Vries. Die mutationstheorie. 2 vols. Leipsic: Veit & Co, 1903, 1901.
[30]
Jeffery P Demuth, Tijl De Bie, Jason E Stajich, Nello Cristianini, and
Matthew W Hahn. The evolution of mammalian gene families. PLOS ONE,
1(1):e85, 2006.
[31]
Jeffery P Demuth and Matthew W Hahn. The life and death of gene families.
Bioessays, 31(1):29–39, 2009.
[32]
Hamsa Dhwani Tadepally and Muriel Aubry. Evolution of C2H2 zinc-finger
gene families in mammals. eLS, 2010.
[33]
Guohui Ding, Peter Lorenz, Michael Kreutzer, Yixue Li, and Hans-Juergen
Thiesen. Sysznf: the c2h2 zinc finger gene database. Nucleic acids research,
37(suppl 1):D267–D273, 2009.
[34]
Jean-Philippe Doyon, Celine Scornavacca, K Yu Gorbunov, Gergely J Szöllősi,
Vincent Ranwez, and Vincent Berry. An efficient algorithm for gene/species
trees parsimonious reconciliation with losses, duplications and transfers. In
Comparative genomics, pages 93–108. Springer, 2011.
[35]
John W Drake, Brian Charlesworth, Deborah Charlesworth, and James F
Crow. Rates of spontaneous mutation. Genetics, 148(4):1667–1686, 1998.
[36]
Alexei J Drummond and Andrew Rambaut. BEAST: Bayesian evolutionary
analysis by sampling trees. BMC evolutionary biology, 7(1):214, 2007.
48
BIBLIOGRAPHY
[37]
Dannie Durand, Bjarni V Halldórsson, and Benjamin Vernot. A hybrid micromacroevolutionary approach to gene tree reconstruction. Journal of Computational Biology, 13(2):320–335, 2006.
[38]
Anthony WF Edwards and Cavalli LL Sforza. The reconstruction of evolution.
Heredity, 18, 1963.
[39]
Greg Elgar and Tanya Vavouri. Tuning in to the signals: noncoding sequence
conservation in vertebrate genomes. Trends in genetics, 24(7):344–352, 2008.
[40]
Isaac Elias. Settling the intractability of multiple alignment. Journal of
Computational Biology, 13(7):1323–1339, 2006.
[41]
Isaac Elias and Jens Lagergren. Fast neighbor joining. In Automata, Languages and Programming, pages 1263–1274. Springer, 2005.
[42]
Ryan O Emerson and James H Thomas. Adaptive evolution in zinc finger
transcription factors. PLOS Genetics, 5(1):e1000325, 2009.
[43]
Joseph Felsenstein. Maximum likelihood and minimum-steps methods for
estimating evolutionary trees from data on discrete characters. Systematic
Biology, 22(3):240–249, 1973.
[44]
Joseph Felsenstein. Maximum-likelihood estimation of evolutionary trees from
continuous characters. American journal of human genetics, 25(5):471, 1973.
[45]
Joseph Felsenstein. Cases in which parsimony or compatibility methods will
be positively misleading. Systematic Biology, 27(4):401–410, 1978.
[46]
Joseph Felsenstein. Evolutionary trees from DNA sequences: a maximum
likelihood approach. Journal of molecular evolution, 17(6):368–376, 1981.
[47]
Joseph Felsenstein and Joseph Felenstein. Inferring phylogenies, volume 2.
Sinauer Associates Sunderland, 2004.
[48]
Robert D Finn, Jody Clements, and Sean R Eddy. HMMER web server:
interactive sequence similarity searching. Nucleic acids research, page gkr367,
2011.
[49]
Walter M Fitch. Distinguishing homologous from analogous proteins. Systematic Biology, 19(2):99–113, 1970.
[50]
T Flouri, F Izquierdo-Carrasco, D Darriba, AJ Aberer, L-T Nguyen,
BQ Minh, A Von Haeseler, and A Stamatakis. The phylogenetic likelihood
library. Systematic biology, page syu084, 2014.
[51]
Allan Force, Michael Lynch, F Bryan Pickett, Angel Amores, Yi-lin Yan,
and John Postlethwait. Preservation of duplicate genes by complementary,
degenerative mutations. Genetics, 151(4):1531–1545, 1999.
BIBLIOGRAPHY
49
[52]
Kristoffer Forslund. The relationship between orthology, protein domain architecture and protein function [Elektronisk resurs]. Department of Biochemistry
and Biophysics, Stockholm University, Stockholm, 2011. ISBN 978-91-7447350-6.
[53]
Pauline Funchain, Annie Yeung, Jean Lee Stewart, Rose Lin, Malgorzata M
Slupska, and Jeffrey H Miller. The consequences of growth of a mutator
strain of escherichia coli as measured by loss of function among multiple gene
targets and loss of fitness. Genetics, 154(3):959–970, 2000.
[54]
Olivier Gascuel. BIONJ: an improved version of the NJ algorithm based on
a simple model of sequence data. Molecular biology and evolution, 14(7):
685–695, 1997.
[55]
Antoine Giraud, Michel Fons, and Francois Taddei. [impact of mutation rate
on the adaptation of gut bacteria]. Journal de la Société de biologie, 197(4):
389–396, 2002.
[56]
Morris Goodman, John Czelusniak, G William Moore, AE Romero-Herrera,
and Genji Matsuda. Fitting the gene lineage into its species lineage, a parsimony strategy illustrated by cladograms constructed from globin sequences.
Systematic Biology, 28(2):132–163, 1979.
[57]
J Guex. Environmental stress and atavism in ammonoid evolution. Eclogae
Geologicae Helvetiae, 94(3):321–328, 2001.
[58]
Stéphane Guindon, Jean-François Dufayard, Vincent Lefort, Maria Anisimova, Wim Hordijk, and Olivier Gascuel. New algorithms and methods
to estimate maximum-likelihood phylogenies: assessing the performance of
PhyML 3.0. Systematic biology, 59(3):307–321, 2010.
[59]
Wilfried Haerty and Chris P. Ponting.
No gene in the genome
makes sense except in the light of evolution.
Annual Review
of Genomics and Human Genetics, 15(1):71–92, 2014.
URL
http://dx.doi.org/10.1146/annurev-genom-090413-025621.
PMID:
24773316.
[60]
Matthew W Hahn, Jeffery P Demuth, and Sang-Gook Han. Accelerated rate
of gene gain and loss in primates. Genetics, 177(3):1941–1949, 2007.
[61]
Matthew W Hahn, Mira V Han, and Sang-Gook Han. Gene family evolution
across 12 drosophila genomes. PLOS genetics, 3(11):e197, 2007.
[62]
JBS Haldane. The part played by recurrent mutation in evolution. American
Naturalist, pages 5–19, 1933.
50
BIBLIOGRAPHY
[63]
Mike Hallett, Jens Lagergren, and Ali Tofigh. Simultaneous identification of
duplications and lateral transfers. In Proceedings of the eighth annual international conference on Research in computational molecular biology, pages
347–356. ACM, 2004.
[64]
Aaron T Hamilton, Stuart Huntley, Mary Tran-Gyamfi, Daniel M Baggott,
Laurie Gordon, and Lisa Stubbs. Evolutionary expansion and divergence in
the ZNF91 subfamily of primate-specific zinc finger genes. Genome research,
16(5):584–594, 2006.
[65]
Paul M Harrison, Nathaniel Echols, and Mark B Gerstein. Digging for dead
genes: an analysis of the characteristics of the pseudogene population in the
Caenorhabditis elegans genome. Nucleic acids research, 29(3):818–830, 2001.
[66]
Paul M Harrison, Duncan Milburn, Zhaolei Zhang, Paul Bertone, and Mark
Gerstein. Identification of pseudogenes in the Drosophila melanogaster
genome. Nucleic acids research, 31(3):1033–1037, 2003.
[67]
David Haussler, Stephen J O’Brien, Oliver A Ryder, F Keith Barker, Michele
Clamp, Andrew J Crawford, Robert Hanner, Olivier Hanotte, Warren E Johnson, Jimmy A McGuire, et al. Genome 10k: a proposal to obtain wholegenome sequence for 10 000 vertebrate species. Journal of Heredity, 100(6):
659–674, 2009.
[68]
Joseph Heled and Alexei J Drummond. Bayesian inference of species trees
from multilocus data. Molecular biology and evolution, 27(3):570–580, 2010.
[69]
Jody Hey and Carlos A Machado. The study of structured populations—
new hope for a difficult and divided science. Nature Reviews Genetics, 4(7):
535–543, 2003.
[70]
Francis G Howarth. High-stress subterranean habitats and evolutionary
change in cave-inhabiting arthropods. American Naturalist, pages S65–S77,
1993.
[71]
Katharina T Huber, Bengt Oxelman, Martin Lott, and Vincent Moulton.
Reconstructing the evolutionary history of polyploids from multilabeled trees.
Molecular biology and evolution, 23(9):1784–1791, 2006.
[72]
John P. Huelsenbeck and Fredrik Ronquist. MRBAYES: Bayesian inference
of phylogenetic trees. Bioinformatics, 17(8):754–755, 2001.
[73]
Raymond B Huey and Peter D Ward. Hypoxia, global warming, and terrestrial late permian extinctions. Science, 308(5720):398–401, 2005.
[74]
Stuart Huntley, Daniel M Baggott, Aaron T Hamilton, Mary Tran-Gyamfi,
Shan Yang, Joomyeong Kim, Laurie Gordon, Elbert Branscomb, and Lisa
Stubbs. A comprehensive catalog of human KRAB-associated zinc finger
BIBLIOGRAPHY
51
genes: insights into the evolutionary history of a large family of transcriptional
repressors. Genome research, 16(5):669–677, 2006.
[75]
David Jablonski. Geographic variation in the molluscan recovery from the
end-cretaceous extinction. Science, 279(5355):1327–1330, 1998.
[76]
David Jablonski. Mass extinctions and macroevolution. Paleobiology, 31(sp5):
192–210, 2005. URL http://www.psjournals.org/doi/abs/10.1666/.
[77]
C Jacq, JR Miller, and GG Brownlee. A pseudogene structure in 5S DNA of
xenopus laevis. Cell, 12(1):109–120, 1977.
[78]
Yuannian Jiao, Norman J Wickett, Saravanaraj Ayyampalayam, André S
Chanderbali, Lena Landherr, Paula E Ralph, Lynn P Tomsho, Yi Hu, Haiying Liang, Pamela S Soltis, et al. Ancestral polyploidy in seed plants and
angiosperms. Nature, 473(7345):97–100, 2011.
[79]
David T Jones, William R Taylor, and Janet M Thornton. The rapid generation of mutation data matrices from protein sequences. Computer applications
in the biosciences: CABIOS, 8(3):275–282, 1992.
[80]
TH Jukes and CR Cantor. Evolution of protein molecules pp. 21–132 in HN
Munro, ed. Mammalian protein metabolism, 1969.
[81]
Henrik Kaessmann. Origins, evolution, and phenotypic impact of new genes.
Genome research, 20(10):1313–1326, 2010.
[82]
Olga K Kamneva and Naomi L Ward. Reconciliation approaches to determining HGT, Duplications, and Losses in Gene Trees. Methods in Microbiology,
41:183–199, 2014.
[83]
Elisabeth Kay, Timothy M Vogel, Frank Bertolla, Renaud Nalin, and Pascal Simonet. In situ transfer of antibiotic resistance genes from transgenic
(transplastomic) tobacco plants to bacteria. Applied and environmental microbiology, 68(7):3345–3351, 2002.
[84]
Patrick J Keeling and Jeffrey D Palmer. Horizontal gene transfer in eukaryotic
evolution. Nature Reviews Genetics, 9(8):605–618, 2008.
[85]
David G Kendall. On the generalized" birth-and-death" process. The annals
of mathematical statistics, pages 1–15, 1948.
[86]
Motoo Kimura. A simple method for estimating evolutionary rates of base
substitutions through comparative studies of nucleotide sequences. Journal
of molecular evolution, 16(2):111–120, 1980.
52
BIBLIOGRAPHY
[87]
Fyodor A Kondrashov and Alexey S Kondrashov. Measurements of spontaneous rates of mutations in the recent past and the near future. Philosophical Transactions of the Royal Society B: Biological Sciences, 365(1544):
1169–1176, 2010.
[88]
Sudhir Kumar, Alan J Filipski, Fabia U Battistuzzi, Sergei L Kosakovsky
Pond, and Koichiro Tamura. Statistics and truth in phylogenomics. Molecular
biology and evolution, page msr202, 2011.
[89]
Nicolas Lartillot and Hervé Philippe. A Bayesian mixture model for acrosssite heterogeneities in the amino-acid replacement process. Molecular biology
and evolution, 21(6):1095–1109, 2004.
[90]
Si Quang Le and Olivier Gascuel. An improved general amino acid replacement matrix. Molecular biology and evolution, 25(7):1307–1320, 2008.
[91]
Liang Liu and Dennis K Pearl. Species trees from gene trees: reconstructing
bayesian posterior distributions of a species phylogeny using estimated gene
tree distributions. Systematic Biology, 56(3):504–514, 2007.
[92]
Liang Liu, Lili Yu, Dennis K Pearl, and Scott V Edwards. Estimating species
phylogenies using coalescence times among sequences. Systematic Biology, 58
(5):468–477, 2009.
[93]
Camilla Looman, Magnus Åbrink, Charlotta Mark, and Lars Hellman. Krab
zinc finger proteins: an analysis of the molecular mechanisms governing their
increase in numbers and complexity during evolution. Molecular biology and
evolution, 19(12):2118–2130, 2002.
[94]
Michael Lynch and John S Conery. The evolutionary fate and consequences
of duplicate genes. Science, 290(5494):1151–1155, 2000.
[95]
Owais Mahmudi, Joel Sjöstrand, Bengt Sennblad, and Jens Lagergren.
Genome-wide probabilistic reconciliation analysis across vertebrates. BMC
Bioinformatics, 14(Suppl 15):S10, 2013.
[96]
Ana C Marques, Jennifer Tan, Sheena Lee, Lesheng Kong, Andreas Heger,
and Chris P Ponting. Evidence for conserved post-transcriptional roles of unitary pseudogenes and for frequent bifunctionality of mrnas. Genome Biology,
13(11):R102, 2012.
[97]
Tomas Marques-Bonet, Santhosh Girirajan, and Evan E Eichler. The origins
and impact of primate segmental duplications. Trends in Genetics, 25(10):
443–454, 2009.
[98]
Barbara McClintock. The origin and behavior of mutable loci in maize. Proceedings of the National Academy of Sciences, 36(6):344–355, 1950.
BIBLIOGRAPHY
[99]
53
Gregor Mendel. Experiments in plant hybridisation. Cosimo, Inc., 2008.
[100] Chen Meng and Laura Salter Kubatko. Detecting hybrid speciation in the
presence of incomplete lineage sorting using gene tree incongruence: a model.
Theoretical population biology, 75(1):35–45, 2009.
[101] Axel Meyer and Yves Van de Peer. From 2r to 3r: evidence for a fish-specific
genome duplication (fsgd). Bioessays, 27(9):937–945, 2005.
[102] HJ Muller. The origination of chromatin deficiencies as minute deletions
subject to insertion elsewhere. Genetica, 17(3):237–252, 1935.
[103] Enrique M Muro, Nancy Mah, Gabriel Moreno-Hagelsieb, and Miguel A
Andrade-Navarro. The pseudogenes of Mycobacterium leprae reveal the functional relevance of gene order within operons. Nucleic acids research, page
gkq1067, 2010.
[104] Saul B Needleman and Christian D Wunsch. A general method applicable to
the search for similarities in the amino acid sequence of two proteins. Journal
of molecular biology, 48(3):443–453, 1970.
[105] Rasmus Nielsen, Carlos Bustamante, Andrew G Clark, Stephen Glanowski,
Timothy B Sackton, Melissa J Hubisz, Adi Fledel-Alon, David M Tanenbaum,
Daniel Civello, Thomas J White, et al. A scan for positively selected genes
in the genomes of humans and chimpanzees. PLOS Biology, 3(6):e170, 2005.
[106] Rasmus Nielsen and Ziheng Yang. Likelihood models for detecting positively
selected amino acid sites and applications to the hiv-1 envelope gene. Genetics, 148(3):929–936, 1998.
[107] Yoshihito Niimura and Masatoshi Nei. Extensive gains and losses of olfactory
receptor genes in mammalian evolution. PLOS ONE, 2(8):e708, 2007.
[108] Cédric Notredame, Desmond G Higgins, and Jaap Heringa. T-coffee: A
novel method for fast and accurate multiple sequence alignment. Journal of
molecular biology, 302(1):205–217, 2000.
[109] Katja Nowick, Miguel Carneiro, and Rui Faria. A prominent role of krab-znf
transcription factors in mammalian speciation? Trends in Genetics, 29(3):
130–139, 2013.
[110] Katja Nowick, Aaron T Hamilton, Huimin Zhang, and Lisa Stubbs. Rapid
sequence and expression divergence suggest selection for novel function in
primate-specific krab-znf genes. Molecular biology and evolution, 27(11):2606–
2617, 2010.
[111] Susumu Ohno et al. Evolution by gene duplication. London: George Alien &
Unwin Ltd. Berlin, Heidelberg and New York: Springer-Verlag., 1970.
54
BIBLIOGRAPHY
[112] Richard Owen. Lectures on the comparative anatomy and physiology of the
invertebrate animals. Longman, 1855.
[113] Heidi G Parker, Bridgett M VonHoldt, Pascale Quignon, Elliott H Margulies,
Stephanie Shao, Dana S Mosher, Tyrone C Spady, Abdel Elkahloun, Michele
Cargill, Paul G Jones, et al. An expressed fgf4 retrogene is associated with
breed-defining chondrodysplasia in domestic dogs. Science, 325(5943):995–
998, 2009.
[114] Paul B Rainey. Evolutionary genetics: the economics of mutation. Current
biology, 9(10):R371–R373, 1999.
[115] Bruce Rannala and Ziheng Yang. Bayes estimation of species divergence
times and ancestral population sizes using DNA sequences from multiple loci.
Genetics, 164(4):1645–1656, 2003.
[116] Matthew D Rasmussen and Manolis Kellis. Accurate gene-tree reconstruction by learning gene-and species-specific substitution rates across multiple
complete genomes. Genome research, 17(12):1932–1942, 2007.
[117] Matthew D Rasmussen and Manolis Kellis. A Bayesian approach for fast and
accurate gene tree reconstruction. Molecular Biology and Evolution, 28(1):
273–290, 2011.
[118] Matthew D Rasmussen and Manolis Kellis. Unified modeling of gene duplication, loss, and coalescence using a locus tree. Genome research, 22(4):
755–765, 2012.
[119] David M Raup. The role of extinction in evolution. Proceedings of the National
Academy of Sciences, 91(15):6758–6763, 1994.
[120] Noah A Rosenberg and Magnus Nordborg. Genealogical trees, coalescent
theory and the analysis of genetic polymorphisms. Nature Reviews Genetics,
3(5):380–390, 2002.
[121] Harald S Rothenfluh, Robert V Blanden, and Edward J Steele. Evolution of V
genes: DNA sequence structure of functional germline genes and pseudogenes.
Immunogenetics, 42(3):159–171, 1995.
[122] Naruya Saitou and Masatoshi Nei. The neighbor-joining method: a new
method for reconstructing phylogenetic trees. Molecular biology and evolution,
4(4):406–425, 1987.
[123] Frederick Sanger, Steven Nicklen, and Alan R Coulson. Dna sequencing with
chain-terminating inhibitors. Proceedings of the National Academy of Sciences, 74(12):5463–5467, 1977.
BIBLIOGRAPHY
55
[124] Claudine Schiff, Michele Milili, and Michel Fougereau. Functional and pseudogenes are similarly organized and may equally contribute to the extensive
antibody diversity of the IgVHII family. The EMBO journal, 4(5):1225, 1985.
[125] Bengt Sennblad. Extending models of genome evolution to include nonhierarchical events. In ESEB XI Congress, Uppsala, Sweden, August 20-25,
2007 (oral presentation), 2007.
[126] Martin Simonsen, Thomas Mailund, and Christian NS Pedersen. Rapid
neighbour-joining. In Algorithms in Bioinformatics, pages 113–122. Springer,
2008.
[127] Kimmen Sjölander. Phylogenomic inference of protein molecular function:
advances and challenges. Bioinformatics, 20(2):170–179, 2004.
[128] Joel Sjöstrand. Reconciling gene family evolution and species evolution. 2013.
[129] Joel Sjöstrand, Bengt Sennblad, Lars Arvestad, and Jens Lagergren. DLRS:
gene tree evolution in light of a species tree. Bioinformatics, 28(22):2994–
2995, 2012.
[130] Joel Sjöstrand, Ali Tofigh, Vincent Daubin, Lars Arvestad, Bengt Sennblad,
and Jens Lagergren. A Bayesian Method for Analyzing Lateral Gene Transfer.
Systematic Biology, 2014.
[131] Temple F Smith and Michael S Waterman. Identification of common molecular subsequences. Journal of molecular biology, 147(1):195–197, 1981.
[132] Alexandros Stamatakis. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics,
22(21):2688–2690, 2006.
[133] Arlin Stoltzfus. On the possibility of constructive neutral evolution. Journal
of Molecular Evolution, 49(2):169–181, 1999.
[134] Marc A Suchard. Stochastic models for horizontal gene transfer taking a
random walk through tree space. Genetics, 170(1):419–431, 2005.
[135] Marc A Suchard and Benjamin D Redelings. BAli-Phy: simultaneous
Bayesian inference of alignment and phylogeny. Bioinformatics, 22(16):2047–
2048, 2006.
[136] Ralf Sudbrak, Richard Reinhardt, Steffen Hennig, Hans Lehrach, Eberhard
Günther, and Lutz Walter. Comparative and evolutionary analysis of the
rhesus macaque extended MHC class II region. Immunogenetics, 54(10):699–
704, 2003.
56
BIBLIOGRAPHY
[137] Örjan Svensson, Lars Arvestad, and Jens Lagergren. Genome-wide survey for
biologically functional pseudogenes. PLOS computational biology, 2(5):e46,
2006.
[138] D. L. Swofford. PAUP*. Phylogenetic Analysis Using Parsimony (*and Other
Methods). Version 4. Sinauer Associates, Sunderland, Massachusetts, 2003.
[139] Gergely J Szöllősi, Bastien Boussau, Sophie S Abby, Eric Tannier, and Vincent Daubin. Phylogenetic modeling of lateral gene transfer reconstructs
the pattern and relative timing of speciations. Proceedings of the National
Academy of Sciences, 109(43):17513–17518, 2012.
[140] Hamsa D Tadepally, Gertraud Burger, and Muriel Aubry. Evolution of C2H2zinc finger genes and subfamilies in mammals: species-specific duplication and
loss of clusters, genes and effector domains. BMC evolutionary biology, 8(1):
176, 2008.
[141] Koichiro Tamura, Daniel Peterson, Nicholas Peterson, Glen Stecher,
Masatoshi Nei, and Sudhir Kumar. MEGA5: molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum
parsimony methods. Molecular biology and evolution, 28(10):2731–2739, 2011.
[142] James H Thomas and Ryan O Emerson. Evolution of c2h2-zinc finger genes
revisited. BMC evolutionary biology, 9(1):51, 2009.
[143] A Tofigh. Using trees to capture reticulate evolution. PhD Thesis, KTH
School of Computer Science and Communication., 2009.
[144] Ali Tofigh, Michael Hallett, and Jens Lagergren. Simultaneous identification of duplications and lateral gene transfers. IEEE/ACM Transactions on
Computational Biology and Bioinformatics (TCBB), 8(2):517–535, 2011.
[145] Benjamin Vernot, Maureen Stolzer, Aiton Goldman, and Dannie Durand.
Reconciliation with non-binary species trees. Journal of Computational Biology, 15(8):981–1006, 2008.
[146] Ilan Wapinski, Avi Pfeffer, Nir Friedman, and Aviv Regev. Automatic
genome-wide reconstruction of phylogenetic gene trees. Bioinformatics, 23
(13):i549–i558, 2007.
[147] Ilan Wapinski, Avi Pfeffer, Nir Friedman, and Aviv Regev. Natural history
and evolutionary principles of gene duplication in fungi. Nature, 449(7158):
54–61, 2007.
[148] James D Watson and Francis HC Crick. The structure of dna. In Cold Spring
Harbor Symposia on Quantitative Biology, volume 18, pages 123–131. Cold
Spring Harbor Laboratory Press, 1953.
BIBLIOGRAPHY
57
[149] Franz Weiling. Historical study: Johann Gregor Mendel 1822–1884. American
journal of medical genetics, 40(1):1–25, 1991.
[150] August Weismann, Edward Bagnall Poulton, Selmar Schönland, and
Arthur Everett Shipley. Essays upon heredity and kindred biological problems, volume 1. Clarendon press, 1891.
[151] Jia F Weng, Doreen A Thomas, and Iven Mareels. Maximum parsimony, substitution model, and probability phylogenetic trees. Journal of Computational
Biology, 18(1):67–80, 2011.
[152] Mary Jane West-Eberhard. Developmental plasticity and evolution. Oxford
University Press, 2003.
[153] Gane K. Wong and Michael Deyholos.
1000 plant genomes project.
"http://www.biology.ualberta.ca/gane_wong_lab/Research.html",
2008. Accessed: 2015-03-09.
[154] Ray Wu and Ellen Taylor. Nucleotide sequence analysis of dna: Ii. complete
nucleotide sequence of the cohesive ends of bacteriophage λ dna. Journal of
molecular biology, 57(3):491–511, 1971.
[155] Yi-Chieh Wu, Matthew D Rasmussen, Mukul S Bansal, and Manolis Kellis.
Most parsimonious reconciliation in the presence of gene duplication, loss,
and deep coalescence using labeled coalescent trees. Genome research, 24(3):
475–486, 2014.
[156] Ziheng Yang. PAML: a program package for phylogenetic analysis by maximum likelihood. CABIOS, 13(5):555–556, 1997.
[157] Yun Yu, R Matthew Barnett, and Luay Nakhleh. Parsimonious inference
of hybridization in the presence of incomplete lineage sorting. Systematic
biology, page syt037, 2013.
[158] Yun Yu, James H Degnan, and Luay Nakhleh. The probability of a gene tree
topology within a phylogenetic network with applications to hybridization
detection. PLOS Genetics, 8(4):e1002660, 2012.
[159] Yun-wu Zhang, Shijie Liu, Xue Zhang, Wu-Bo Li, Yaomin Chen, Xiumei
Huang, Liangwu Sun, Wenjie Luo, William J Netzer, Richard Threadgill,
et al. A Functional Mouse Retroposed Gene Rps23r1 Reduces Alzheimer’s
β-Amyloid Levels and Tau Phosphorylation. Neuron, 64(3):328–340, 2009.
[160] ZhaoLei Zhang and Mark Gerstein. Large-scale analysis of pseudogenes in the
human genome. Current opinion in genetics & development, 14(4):328–335,
2004.
58
BIBLIOGRAPHY
[161] Olga Zhaxybayeva and W Ford Doolittle. Lateral gene transfer. Current
Biology, 21(7):R242–R246, 2011.
[162] Deyou Zheng, Adam Frankish, Robert Baertsch, Philipp Kapranov, Alexandre Reymond, Siew Woh Choo, Yontao Lu, France Denoeud, Stylianos E
Antonarakis, Michael Snyder, et al. Pseudogenes in the encode regions: consensus annotation, analysis of transcription, and evolution. Genome research,
17(6):839–851, 2007.
[163] Yi Zhou and Bud Mishra. Quantifying the mechanisms for segmental duplications in mammalian genomes by statistical analysis and modeling. Proceedings
of the National Academy of Sciences of the United States of America, 102(11):
4051–4056, 2005.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement