Machine learning approach to decipher the genetic architecture of oligo- to polygenic disorders

Time frame: 
October 1, 2013 to October 1, 2015

Recent advances in genome-wide sequencing technologies and data analysis have proven to be very successful in the delineation of the genetic basis of hitherto poorly understood, fully penetrant, monogenic disorders. In practice, disease-causing genetic variants are identified through a set of filtering steps, which potentially reduce the large collection of exonic variants, typically around 20000, to one or a few variants (or genes).  These filtering steps eliminate variants by assuming that the causative variants alter the RNA splicing or protein sequence, that they are rare in the normal healthy population, show complete penetrance and detectance. Although significant successes have been achieved with this approach, additional tools are still required when the outcome of the variant filtering does not provide a simple answer.

The task of filtering out the causative variants/genes becomes even more challenging when dealing with congenital conditions with complex inheritance patterns, distributed along an oligo- to polygenic continuum. In this context, the assumption is that a combination of rare variants spread over protein-coding and microRNA genes is causative for the observed phenotype. Since these variants are rare in the population, it makes them hard to detect using standard single variant association tests and therefore more aggregative techniques are required. In this project we aim to develop methods that can distill, from the large collection of rare variants, those that have any relevance to the phenotype under investigation.

Promoters: Guillaume Smits, Sonia Van Dooren and Tom Lenaerts

Co-Promoters: Marianne Rooman, Didier Croes, Catheline Vilain, Nicolas Simonis and Wim Vranken

Partners: M. Abramowicz, M. Bonduelle, Ben Caljon, F. Libert, S. Seneca, W. Lissens, M. Prevost and P. Tompa