Novel Scalable Approaches for Multiple Sequence Alignment and Phylogenomic Reconstruction

360 Pages·2015·6.87 MB·English

Copyright
by
Siavash Mir arabbaygi
2015
The Dissertation Committee for Siavash Mir arabbaygi
certifies that this is the approved version of the following dissertation:
Novel scalable approaches for multiple sequence
alignment and phylogenomic reconstruction
Committee:
Keshav Pingali, Supervisor
Tandy Warnow, Co-Supervisor
David Hillis
Bonnie Berger
Joydeep Ghosh
Ray Mooney
Novel scalable approaches for multiple sequence
alignment and phylogenomic reconstruction
by
Siavash Mir arabbaygi, B.S.; M. APPL S.
DISSERTATION
Presented to the Faculty of the Graduate School of
The University of Texas at Austin
in Partial Fulfillment
of the Requirements
for the Degree of
DOCTOR OF PHILOSOPHY
THE UNIVERSITY OF TEXAS AT AUSTIN
August 2015
Dedicated to my mother, and the memory of my father.
Acknowledgments
I wish to thank the countless people who helped me throughout my
PhD. I enjoyed working with many collaborators, most of whom I never met,
and not all of whom I can name here. I had the good fortune of working with
my labmates, especially Nam Nguyen, Shamsuzzoha Md. Bayzid, and Th´eo
Zimmermann, and our interactions have always been smooth and fruitful. I
benefited tremendously from my involvement in both the avian phylogenomics
and the thousand transcripts projects, both of which included many top re-
searchers from across the world. I want to especially thank Bastien Boussau,
Edward Braun, Tom Gilbert, Erich Jarvis, Jim Leebens-Mack, Gane Wong,
Norman Wickett, and Guojie Zhang. I also thank Kevin Liu, Mark Holder,
and Jeet Sukumaran for allowing me to use their code throughout my work. I
would also like to thank all members of my committee. My understanding of
the biology was tremendously improved by participation in discussions lead by
Prof. David Hillis. I am also grateful to Prof. Keshav Pingali, who accepted
me in his group after my original supervisor officialy left UT; Keshav was a
role model, and involvement with his group has greatly expanded my depth.
I should also thank my sources of funding (Howard Hughes Medical Insti-
tute international student fellowship, Canadian NSERC Doctoral award, and
NSF) and the generous computational support of Texas Advanced Computing
Center (TACC) and the computing cluster (Condor) at the CS department.
v
I cannot start to thank enough my supervisor, Prof. Tandy Warnow,
for all her guidance and her support. I enjoyed the chance to argue with her
frequently on all matters related to our research, small or large, and even
though I often found myself on the loosing side of these arguments, I always
felt encouraged. Nothing like her openness and acceptance in rare occasions I
won an argument could have better trained me for the skepticism and inquis-
itiveness necessary for the scientific pursuit. I can only inspire to match her
enthusiasm for research, which was a constant boost of energy to me through-
out. Her endless support as a research advisor and a career mentor has opened
to me many new doors, which I did not believe accessible when I started my
studies. I owe much of my past and future achievements to her.
For all the support I had at work, no progress was possible if it wasn’t
for the persistent emotional support my fiancee, now wife, showered me with.
She raised my spirits with expressions of confidence in my abilities, warranted
or not. Her confidence in me was a constant motivation to overcome obstacles.
And that’s to say nothing of all the day-to-day ways in which she has nudged
me towards being a more organized and focused student. I would have never
beenintheofficeearlyinthemorningeverydayoftheweek,ifitwasn’tforher.
She was a major reason I started my PhD studies, and her encouragements
are a major reason I feel confident continuing as a researchers.
Finally, everything I have ever achieved has been possible only because
of my mother, who thought me starting is easy, persevering hard, and letting
go the hardest. Her example is always with me.
vi
Novel scalable approaches for multiple sequence
alignment and phylogenomic reconstruction
Publication No.
Siavash Mir arabbaygi, Ph.D.
The University of Texas at Austin, 2015
Supervisors: Keshav Pingali
Tandy Warnow
The amount of biological sequence data is increasing rapidly, a promis-
ing development that would transform biology if we can develop methods that
can analyze large-scale data efficiently and accurately. A fundamental ques-
tion in evolutionary biology is building the tree of life: a reconstruction of
relationships between organisms in evolutionary time. Reconstructing phy-
logenetic trees from molecular data is an optimization problem that involves
many steps. In this dissertation, we argue that to answer long-standing phylo-
genetic questions with large-scale data, several challenges need to be addressed
in various steps of the pipeline. One challenges is aligning large number of se-
quences so that evolutionarily related positions in all sequences are put in the
same column. Constructing alignments is necessary for phylogenetic recon-
struction, but also for many other types of evolutionary analyses. In response
to this challenge, we introduce PASTA, a scalable and accurate algorithm that
vii
can align datasets with up to a million sequences. A second challenge is re-
lated to the interesting fact that various parts of the genome can have different
evolutionary histories. Reconstructing a species tree from genome-scale data
needs to account for these differences. A main approach for species tree re-
construction is to first reconstruct a set of “gene trees” from different parts of
the genome, and to then summarize these gene trees into a single species tree.
We argue that this approach can suffer from two challenges: reconstruction of
individual gene trees from limited data can be plagued by estimation error,
which translates to errors in the species tree, and also, methods that sum-
marize gene trees are not scalable or accurate enough under some conditions.
To address the first challenge, we introduce statistical binning, a method that
re-estimates gene trees by grouping them into bins. We show that binning
improves gene tree accuracy, and consequently the species tree accuracy. To
address the second challenge, we introduce ASTRAL, a new summary method
that can run on a thousand genes and a thousand species in a day and has
outstanding accuracy. We show that the development of these methods has
enabled biological analyses that were otherwise not possible.
viii
Table of Contents
Acknowledgments v
Abstract vii
List of Tables xv
List of Figures xvi
Chapter 1. Introduction 1
Chapter 2. Background 11
2.1 Phylogeny: an evolutionary tree . . . . . . . . . . . . . . . . . 11
2.1.1 Properties of a phylogenetic tree . . . . . . . . . . . . . 12
2.1.2 Character evolution . . . . . . . . . . . . . . . . . . . . 15
2.1.2.1 Substitutions . . . . . . . . . . . . . . . . . . . 16
2.1.2.2 Alignments . . . . . . . . . . . . . . . . . . . . 19
2.2 Gene trees and the species tree . . . . . . . . . . . . . . . . . . 21
2.2.1 Definitions and concepts . . . . . . . . . . . . . . . . . . 21
2.2.2 Coalescence and Incomplete Lineage Sorting (ILS) . . . 23
2.2.2.1 Coalescence . . . . . . . . . . . . . . . . . . . . 25
2.2.2.2 Multi-Species Coalescent (MSC) . . . . . . . . . 28
2.3 Phylogenetic reconstruction . . . . . . . . . . . . . . . . . . . 31
2.3.1 Multiple Sequence Alignment (MSA) . . . . . . . . . . . 34
2.3.2 Tree reconstruction . . . . . . . . . . . . . . . . . . . . 37
2.3.2.1 Maximum parsimony . . . . . . . . . . . . . . . 38
2.3.2.2 Maximum likelihood . . . . . . . . . . . . . . . 38
2.3.2.3 Bayesian estimation . . . . . . . . . . . . . . . 39
2.3.2.4 Distance-based . . . . . . . . . . . . . . . . . . 40
2.3.2.5 Branch support . . . . . . . . . . . . . . . . . . 40
ix
2.3.3 Analyzing multiple genes . . . . . . . . . . . . . . . . . 41
2.3.3.1 Concatenation . . . . . . . . . . . . . . . . . . . 42
2.3.3.2 Summary methods . . . . . . . . . . . . . . . . 44
2.3.3.3 Co-estimation of gene trees and species trees . . 46
2.4 Method evaluation . . . . . . . . . . . . . . . . . . . . . . . . 48
2.4.1 Comparing two phylogenies . . . . . . . . . . . . . . . . 49
2.4.2 Comparing two alignments . . . . . . . . . . . . . . . . 50
Chapter3. PASTA:PracticalAlignmentsusingSAT´eandTrAn-
sitivity 52
3.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2 Background: SAT´e-II . . . . . . . . . . . . . . . . . . . . . . . 55
3.3 PASTA’s algorithm . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3.1 Transitivity merge of two alignments . . . . . . . . . . . 58
3.3.2 Steps of a PASTA iteration . . . . . . . . . . . . . . . . 60
3.3.2.1 Six PASTA Steps . . . . . . . . . . . . . . . . . 60
3.3.2.2 Computing the transitivity merge . . . . . . . . 65
3.3.3 Running time . . . . . . . . . . . . . . . . . . . . . . . . 69
3.4 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . 73
3.4.1 Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.4.1.1 Nucleotide . . . . . . . . . . . . . . . . . . . . . 74
3.4.1.2 Amino-acid . . . . . . . . . . . . . . . . . . . . 76
3.4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.4.3 Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.4.4 Computational platform . . . . . . . . . . . . . . . . . . 79
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.5.1 Ability to complete analyses . . . . . . . . . . . . . . . 80
3.5.2 Results on nucleotide datasets . . . . . . . . . . . . . . 81
3.5.3 Alignment accuracy on AA datasets . . . . . . . . . . . 87
3.5.4 Comparisons on larger datasets . . . . . . . . . . . . . . 88
3.5.5 Running Time . . . . . . . . . . . . . . . . . . . . . . . 89
3.5.6 Impact of varying algorithmic parameters. . . . . . . . . 91
3.5.6.1 Starting tree . . . . . . . . . . . . . . . . . . . . 91
x