Compact Course 54

Duration

Tuesday, 23 September 2008 - Wednesday, 24 September 2008

Taught by

Prof. Dr. G. Gonnet

Dr. Maria Anisimova

Abstract

Most recent genomic-scale studies agree that Darwinian selection plays a dominant role in shaping the genomes of living organisms. Using latest statistical techniques, several human genes were reported to evolve adaptively, with some associated to human disease. Studying evolution of protein-coding genes across species also enables us to deduce properties of ancestral proteins, as demonstrated in recent studies of visual (opsin) genes, where dinosaur opsins were deduced statistically and then recreated experimentally. Selective pressure on the protein-coding genes is of a particular interest in the bio-medical research and vaccine development. For example, in viral proteins identifying residues that diversified in samples from infected patients helps to locate amino acids evolving under selective pressure to evade the immune response.

Current statistical models of codon evolution and classical hypothesis testing techniques facilitate robust studies of selection on the coding-protein genes, test theories of gene family diversification, pathogeneicity, viral evolution, predict epidemiological scenarios and estimate the beginning of the epidemics dynamics. The same rigorous framework for statistical testing can also be universally applied to fundamental questions in classical biology and molecular evolution.

We discuss the biological and medical implications on the examples of several recent case studies, and the available software. We review the best statistical methodology for robust inferences and predictions of sites under positive selection, based on the alignments of homologous protein-coding DNA sequences.

We focus on Markovian codon models and tests for positive selection based on such models. Examples from high-profile publications are used to illustrate the applicability and the utility of such techniques throughout the course.

The course will be held in English.

Additional Information

Academic participants are offered a 50% discount.

Program

23.9.2008: Theory session

1) Brief introduction to comparative genomics: the genomics data, sources and availability, gene and homology prediction, sequence alignment. The goals of comparative genomics.

2) Modeling protein-coding sequence evolution: genetic code, Markov models, and common assumptions. Is it worth it?

3) The fundamental concepts of maximum likelihood and classical hypothesis testing: the bare essentials. The neutral theory vs. selectionism. Evaluating the selective pressure on the protein by maximum likelihood.

4) The basics of the Bayesian approach: basic concepts and examples. Application to the prediction of sites under positive selection

5) The genomic scan for candidate genes under positive selection. Reliability, limitations, examples, and conclusions

6) The best recent case studies reporting positive selection. Codon models and epidemiology.

 

24.9.2008: Practical session in the computer laboratory

Hands-on experience in analyzing protein–coding data for adaptive signatures:

1) Detecting positive selection with site and branch-site models, predicting sites under positive selection (with PAML package)

2) Evaluating synonymous rate variation (with HYPHY package)

3) Large-scale automatic analyses (standard Perl scripts provided)

4) Exploring codon adaptation (with DARWIN programming environment)

Goal

• To facilitate the conceptual understanding of the classic statistical methodology used to study the selective signature in genes, gene families, or in full genomes.
• To provide an overview of the best public resources and statistical techniques for natural selection studies.

Target groups

• Specialists and advanced students from the biomedical sciences, computing sciences and bioinformatics
• Anyone interested in statistical and computational applications in genomics and bioinformatics

Prerequisites

The course is designed in a popular style to cater for mixed interdisciplinary audience and so assumes no prior knowledge of modeling or in-depth biology, but will require:

• Interest in statistical analysis of genomic data
• Basic understanding of the probability concept
• Desire to learn about molecular evolution, natural selection and major statistical modeling techniques in simple terms