Date: 15.09, Place: Quantum M1, Time: 12:30
Presentation by: Maiju Pesonen, Post doctoral researcher
Aalto University / Helsinki Institute for Information Technology
Finnish Centre of Excellence in Computational Inference (COIN)
Are they talking to each other or not? – A journey to genome wide modelling of epistasis
In early models of natural selection presented in the early 20th century, each gene was considered to make its own characteristic contribution to the phenotype, against an average background of other genes. However, the understanding has increased considerably through the history of genetics, and today it well known, that genes rarely operate independently from other genes. Epistasis describes how these possibly complex interactions between genes or within a gene can affect phenotypes. Thus, in numerous organisms, epistatic interactions between single nucleotide polymorphisms (SNPs) have a large influence on the capacity of a phenotype trait for adaptive evolution. However, the complexity of the possible set of interactions has mostly limited previous analyses of epistasis to viral datasets of limited diversity. Recent advances in the scale and diversity of population genomic datasets for bacteria now provide the potential for genome-wide patterns of co-evolution to be studied at the resolution of individual bases. Here we describe a new statistical method, genome DCA, able to identify the polymorphic loci from bacterial genome sequence alignments under the strongest co-evolutionary pressures. By considering the evolution of polymorphic sites simultaneously and using the inference tools for regularized statistical model learning one avoids both the problems that drive high levels of false positives and negatives when the number of pairwise interactions grows. The methods introduced here may offer a powerful alternative to traditional genome wide association analyses for multiple unknown phenotypes in the emerging era of massive population sequencing for bacteria.
All interested are welcome!