Section: New Results

In silico evolution improves statistical models of genome dynamics

Using Aevol, we have proved that statistical frameworks published in the last twenty years for inferring evolutionary genome rearrangements are flawed in two ways. First, they mistranslated a null hypothesis on a uniform breakage model, and second, they assumed that genomic breakable regions are known a priori. We propose ways to correct these flaws by combining mathematical approaches, simulations, observations and validation on real genomic data. The results will be of interest for an audience from evolutionary biology, computational biology, bioinformatics and mathematics. We successively show that:

• a truly uniform hypothesis on rearrangement breakages leads to a model with an equilibrium intergene size distribution that fits the measured one on diverse genomes,

• estimations based on the flawed uniform breakage model completely fail on simulations with the truly uniform model,

• coherently with previous studies the flawed, and to a lesser extent, the truly uniform model are rejected on amniote genomes if breakable regions are identified with intergenic regions,

• co-estimating the number of breakable regions with the rearrangement distance gives coherent values on amniote genomes.

A paper reporting these results has been submitted by Priscila Biller, Carole Knibbe and Eric Tannier.