Computational approaches for analyzing the role of protein-DNA interactions in gene regulation [Elektronische Ressource] / vorgelegt von Annette Höglund

Sujets

• Biologie

Informations

• Publié par: eberhard_karls_universitat_tubingen
• Publié le: 01 janvier 2006
• Nombre de lectures: 40
• Langue: English
• Poids de l'ouvrage: 1 Mo

Extrait

Computational Approaches for Analyzing
the Role of Protein-DNA Interactions in
Gene Regulation






Dissertation
der Fakultät für Informations- und Kognitionswissenschaften
der Eberhard-Karls-Universität Tübingen
zur Erlangung des Grades eines
Doktors der Naturwissenschaften
(Dr. rer. nat.)



vorgelegt von
Dipl.-Ing. Annette Höglund
aus Mariehamn, Finnland



Tübingen
2006































Tag der mündlichen Qualifikation: 19.06.2006
Dekan: Prof. Dr. Michael Diehl
1. Berichtstatter: Prof. Dr. Oliver Kohlbacher
2. Berichtstatter: Prof. Dr. Hans-Peter Lenhof
(Universität des Saarlandes)
3. Berichtstatter: Dr. Hagit Shatkay
(Quen’s University, Kingston) Acknowledgements
Standing here with a nice feeling inside, there are several people I want to acknowledge.
First and foremost, I want to thank Prof. Oliver Kohlbacher and Prof. Hans-Peter Lenhof
for giving me the opportunity to choose an interesting research topic and for providing ex-
cellent conditions to explore it both in Tubingen and in Saarbruc ken.
I am forever indebted to Oliver Kohlbacher for an enthusiastic supervision, stimulating
discussions, and for taking me on as a PhD student in the rst place. I admire him for a
sharp mind, a ne sense to combine critique with commitment towards others, and for being
a skilled team leader.
Looking further back in time I want to start with thanking Arne Elofsson at the Stock-
holm Bioinformatics Center, for showing me what computer science can do for life sciences.
Realizing that is my mere motivation for staring at something so incredibly square-shaped
for so incredibly many hours.
I was fortunate to meet Wyeth Wasserman, who warmly welcomed me into his group at the
Karolinska Institute and opened the door to future research. Albin Sandelin, Boris Lenhard,
Nina Stahlberg, and Elena Herzog contributed with inspiring discussions, great work, and a
wonderful group atmosphere.
Thanks to Gene Myers, Vineet Bafna, Samuel Levy, and Sridhar Hannenhalli for inviting
me to do an internship at Celera Genomics. Knut Reinert and Daniel Huson equipped me
with a weapon and protection in order to survive regular afternoon battles. Karen Eilbeck
and Giuseppe Lancia provided housing, dancing company, and were never late for laughs.
Nathan Edwards and Sorin Istrail for infectious mirth. Hagit Shatkay was happy to inherit
my armour and is the most encouraging woman I know. Furthermore, I want to thank Hagit
for her e orts in reviewing the work described in this thesis.
My warm thanks to Hans-Peter Lenhof and the rest of the very friendly Saarbruc ken
kickers, for their hospitality and support during the rst few years in Germany.
Much respect to my current colleagues in Tubingen: Andreas Kerzmann, Pierre D onnes,
Marc Sturm, Torsten Blum, Muriel Quenzer, Jana Schmidt, Nora Toussaint, and Nico Pfeifer
for discussions, thinking, company, care, and assistance with all kinds of technical problems
ivat all times. Thanks to my former and present students for their hard work, inspiration, and
patience. I speci cally want to emphasis the work by Marc Sturm (development and imple-
mentation of CAP), Christine Herold (analysis of human melanoma-related genes), Sebastian
Schultheiss (analysis of stem cell regulation in plants), and Nina Fischer (for her stubbornness
and shared interest in modeling 3D structures). Furthermore, I want to thank the members
of the Zentrum fur Bioinformatik Tubingen (ZBIT) for contributing to enjoyable moments
at Sand and in Oberjoch.
I want to thank Hagit Shatkay, Stefan Rensing, Jan Lohmann, Wolfgang Busch, Hans-
Werner Adolph, Jens Lagergren, Michael Hallett, Michelle Scott, and Martin Latterich for
pleasant cooperations.
I am grateful beyond what words can describe to my parents Ann-Louise and Henry
H oglund for endless encouragement and concern, boosting my self-esteem, and for being
just what they are - my parents. Thanks to my brothers, grandparents, and Pierre’s family
for reminding me about life in general, care, and attention whenever it is needed.
A warm bunch of hugs to all my friends for support, entertaniment, and for trying to keep
me sane. In particular, I want thank to my best friend Johanna Gustafsson for always being
a great source of strength and for kindling an eternal spirit of optimism.
Emmy
our wonderful daughter, for showing me a new dimension to life. Your sparkling smile at
the crack of dawn reminds me of how beautiful life is - every day of it!
Pierre
my everything. You hear my thoughts, understand my dreams, and ll my life with
laughter and love!
vAbstract
Gene regulation plays a pivotal role at all stages of organism development, in cell di eren ti-
ation, and for maintaining homeostasis. Controlled spatial and temporal gene expression is
achieved by means of complex and robust regulatory networks. A key event in maintaining
such networks is the sequence speci c protein-DNA recognition, which enables transcription
factors to identify their respective binding sites.
Computational and structural biologists face intriguing challenges at three di eren t levels
when investigating gene regulation. First, the involvement of gene regulation in disease can
be addressed by studying global e ects of gene regulatory networks, which are visible at the
level of systems. Furthermore, detecting the often short and variable transcription factor
binding sites (TFBSs) in genomic DNA is not a trivial task, since the prediction of TFBSs
and delineation of functional regulatory modules are conducted at the level of sequences.
Finally, there is a challenge in understanding the factors governing transcription factor-DNA
recognition, as the information needs to be collected at the molecular level. Structure-based
methods provide detailed information about protein-DNA interactions at atomic resolution.
In this work, a versatile approach for computational analysis of the di eren t levels of gene
regulation, gradually zooming in from the global level of systems to the molecular level,
is presented. Linking information related to gene regulation from the di eren t levels can
help in clarifying phenomena that are hard to explain using only one source of information.
First, the in uence of gene regulation is analyzed at the level of systems. A set of cancer-
related target genes are identi ed using a novel integrative analysis pipeline. Microarray
data, immunological data, and curated biological knowledge are brought together enabling
extensive analysis of the underlying mechanisms controlling gene expression in cancer tissue.
The transcription factor AP2 is suggested to play a key regulatory role in controlling a set of
over-expressed melanoma-related genes. The computational results presented are supported
by previously reported experimental evidence.
Zooming in to the level of sequences transcription factors orchestrating the expression of
functionally related genes are identi ed in yeast and plant, which are two important model
organisms for studying gene regulation. The pattern- nding algorithm Gibbs sampling is
employed for discovering putative functional TFBSs in functionally related genes. The re-
sponse element ACGCGT is found to be over-represented in DNA-repair genes in yeast,
which supports the idea that the transcription factor MBP1 is involved in blocking repli-
vication of damaged DNA. The vital regulation of stem cells is explored in plant, providing
preliminary computational evidence for TFBSs critical to stem cell di eren tiation.
The nal transition is the step from analyzing gene regulation at the levels of systems and
sequences to studying protein-DNA interactions at atomic detail. Structural data provides an
additional source for gaining insight into the thermodynamic properties of sequence speci c
binding, which eventually directs gene regulation. A computational protocol for analyzing
the e ects that small base modi cations have on the overall binding free energy is described.
The computationally obtained results for mutating the thymine to uracil in transcription
factor-DNA complexes agree well with previously reported experimental results, illustrating
the applicability of the protocol. This is a rst step towards using molecular modeling for
constructing structure-based models of TFBSs.
Each individual level of this step-wise analysis provides crucial information needed to gain
insight into the di eren t aspects underlying complex regulatory control mechanisms. Anal-
ysis at the level of systems and networks is crucial for understanding global e ects of gene
regulation, the implications of gene regulation in disease, and for identifying sets of target
genes. Sequence-based methods are used for discovering functional binding sites in gene regu-
latory regions for such sets of related genes, responsible for directing gene expression. Finally,
structural analysis can explain ambiguities observed in sequence-based models, however, can
only be applied to a limited number of protein-DNA complexes due to high computational