Sequence alignment and phylogenetic inference [Elektronische Ressource] / Roland Fleißner

Sujets

• Biologie

Informations

• Publié par: heinrich-heine-universitat_dusseldorf
• Publié le: 01 janvier 2003
• Nombre de lectures: 8
• Langue: English
• Poids de l'ouvrage: 1 Mo

Extrait

Sequence alignment and
phylogenetic inference
Inaugural-Dissertation
zur
Erlangung des Doktorgrades der
Mathematisch-Naturwissenschaftlichen Fakultat
der Heinrich-Heine-Universitat Dusseldorf
vorgelegt von
Roland Flei ner
aus Munc hen
Dusseldorf
2003Gedruckt mit der Genehmigung der Mathematisch-Naturwissenschaftlichen
Fakult at der Heinrich-Heine-Universit at Dusseldorf
Referent: Prof. Dr. Arndt von Haeseler
Koreferenten: Prof. Dr. Gerhard Steger, Prof. Dr. Rolf Backofen (Jena)
Tag der mundlic hen Prufung: 18. Dezember 2003
iiPremature optimization is the root of all evil.
(Donald Knuth)Acknowledgements
I would like to thank my advisor Arndt von Haeseler for having the idea for
this project and for his patience, Dirk Metzler and Anton Wakolbinger for
their enthusiasm and their expertise, Gunter Wei for explaining basic statis-
tics, Bojan Basrak and Thomas Schlegel for listening to my stories, Heiko
Schmidt and Achim Radtke for giving advice on the usage of computers,
Lutz Voigt for keeping the system running and Tanja Gesell for stimulating
discussions. I also thank everybody at the zoological institute in Munich,
the MPI for evolutionary anthropology in Leipzig, the mathematics depart-
ment in Frankfurt and at the bioinformatics institute in Dusseldorf for the
pleasant working environments. I am indebted to the Deutsche Forschungs-
gemeinschaft and the Max Planck Gesellschaft for nancial support and to
my friends and especially my family for moral back up.
Publications
Parts of this study have already been published or have been submitted for
publication:
1. Roland Flei ner, Dirk Metzler and Arndt von Haeseler. 2000. Can one
estimate distances from pairwise sequence alignments? In: GCB 2000,
Proceedings of the German Conference on Bioinformatics, October 5-7,
2000, Heidelberg (Bornberg-Bauer, E., Rost, U., Stoye, J., and Vingron,
M., eds) pp. 89{95, Logos Verlag Berlin.
2. Dirk Metzler, Roland Flei ner, Anton Wakolbinger and Arndt von Hae-
seler. 2001. Assessing variability by joint sampling of alignments and
mutation rates. J. Mol. Evol. 53:660{669.
ivContents
Introduction 1
1 Basics of phylogenetic inference 3
1.1 Sequence alignment . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Reconstructing phylogenetic trees . . . . . . . . . . . . . . . . 11
2 Distance estimation from pairwise sequence alignments 16
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 The observed distance of optimally aligned sequences . . . . . 19
2.3 A least squares approach to distance estimation . . . . . . . . 29
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3 Reconstructing trees from multiple alignments 35
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Simulation I: The four-taxon case . . . . . . . . . . . . . . . . 37
3.3 Simulation II: Varying gap penalties . . . . . . . . . . . . . . . 39
3.4 Simulation III: In uence on larger trees . . . . . . . . . . . . . 44
3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4 Statistical pairwise alignment 53
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
v4.2 The TKF1 insertion-deletion model . . . . . . . . . . . . . . . 54
4.3 The sampling of pairwise alignments . . . . . . . . . . . . . . 60
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5 Simultaneous statistical multiple alignment and phylogeny
reconstruction 69
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.3 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3.1 Strategy I: pure simulated annealing . . . . . . . . . . 75
5.3.2 Strategy II: simulated annealing with improved start . 80
5.3.3 Strategy III: simulated with improved start
and nearest-neighbor interchanges . . . . . . . . . . . . 81
5.4 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.4.1 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.4.2 An example: HVR-1 from primates . . . . . . . . . . . 86
5.5 Discussion and Outlook . . . . . . . . . . . . . . . . . . . . . 88
6 Summary 89
A Program manuals 91
A.1 ALIFRITZ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
A.2 SIMULATOR . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
B Two useful algorithms 99
B.1 Sampling pairwise alignments in linear space . . . . . . . . . . 99
B.2 Neighbor joining under constraints . . . . . . . . . . . . . . . 105
C Generalizations of the TKF2 model 108
viC.1 Insertion-deletion models with arbitrary distributions of frag-
ment lengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
C.2 models with several classes of fragments . . 112
Bibliography 115
viiIntroduction
When studying a set of related nucleotide or amino acid sequences, we are in-
terested in several questions: What does the underlying tree look like? Which
positions in the sequences are homologous? What are the characteristics of
the evolutionary process that led to these sequences? How reliable are our
deductions? We normally try to tackle these questions by rst constructing
an optimal multiple alignment on which we then base all our inference on
the tree topology and the parameters of the substitution model we prefer.
Due to highly elaborate alignment (e. g. Thompson et al., 1994) and tree re-
construction programs (e. g. Swo ord, 1991; Felsenstein, 1993; Strimmer and
von Haeseler, 1996), this way of proceeding is fast and certainly works well
in most cases. Nevertheless, it has several drawbacks: Firstly, as we did not
attempt to model the insertion and deletion process explicitly, we can hardly
make any statements about the nature of this process. Secondly, in order to
nd an optimal alignment we have to specify mismatch and gap penalties,
thus making implicit assumptions about the plausibility of substitutions, in-
sertions and deletions. Thirdly, multiple alignment methods either align the
sequences according to a prespeci ed phylogenetic tree or ignore their evolu-
tionary history (cf. Gotoh, 1999). Either way, our estimate of the phylogeny
will be a ected. Finally, our knowledge of the statistics of alignment scores
and of the behaviour of the heuristics which are used to come close to the
1optimum is at the moment very limited. Thus, the problem of phylogenetic
analysis is twofold: Our estimates of phylogenies and model parameters are
in uenced by the multiple alignments on which we base our studies, yet we
do not know how to judget errors and thus have no clue about the
reliability of our estimates.
This thesis deals with some of these problems arising at the junction
between sequence alignment and phylogeny reconstruction: After giving a
brief overview over some basic concepts and methods of phylogenetic analysis
in chapter 1, we will study the in uence of global sequence alignment on the
estimation of pairwise distances (chapter 2) as well as on the reconstruction
of phylogentic trees (chapter 3). Chapter 4 gives an introduction into the
application of insertion and deletion models. These models are then used to
simultaneously reconstruct phylogenies and multiple alignments (chapter 5).
Finally, we describe in the appendices the usage and some implementation
details of the main programs that were written for this thesis as well as some
possible generalizations of the models applied.
2Chapter 1
Basics of phylogenetic inference
The problems dealt with in this thesis originate from the study of phyloge-
netic relationships on the basis of biological sequences. In this chapter, we
give a brief introduction into the eld and its terminology. At the center of
all our studies are sets of related biological sequences. These sequences may
be either DNA sequences, proteins or all kinds of RNAs, but for the purpose
of this thesis they are simply thought of as strings whose letters belong to the
canonical alphabets of nucleotides or amino acids (cf. Voet and Voet, 1995).
Thus, we are ignoring any biochemical features of nucleotides or amino acids,
and when we speak of DNA, we will always mean one strand of the double
helix. We consider sequences to be ‘related’ or ‘homologous’ if they derive
from a common ancestor. Due to the way that DNA is replicated, all the
sequences which stem from one ancestral sequence are related by a bifur-
cating tree: One sequence replicates and produces two daughter sequences,
each of these daughter sequences replicates after some time and, in turn,
produces two daughter and so on. Normally we do not have the
whole progeny of an ancestral sequence at hand but only a small subset. Yet,
since picking a subset of the progeny corresponds to pruning the tree, the
3