Distinguishing structure from panmixia: Simona Grusea submitted our latest article to Theoretical Population Biology

The IICR (inverse instantaneous coalescence rate) is time- and sample-dependent parameter that was originally introduced by Olivier Mazet in an article published in 2016 in Heredity. Imagine that you have two haploid genomes, one question a population geneticist would ask is “what could I say with so much genomic data about the demographic history of the population, or of the species these two genomes comes from?“.  Several statistical methods now exist that allow us to answer that question to some extent. For instance, under the assumption of panmixia and absence of population structure we can reconstruct a history of population size changes. This is what several studies have done using the landmark PSMC method of Li and Durbin published in nature in 2011. What Olivier (and colleagues) showed, among other things, is that, if the population sampled was in reality connected to other populations through gene flow and even if it had never changed in size, the inferential methods such as the PSMC would still infer a history of population size changes. He showed that a (spurious) history of population size changes could be predicted exactly and that this (spurious) history would explain the data as well as the real history. In other words, we were facing an identifiability problem: having two genomes was not enough to separate population structure from population size changes. In this new article led by Simona Grusea and Willy Rodriguez, both members of Olivier Mazet’s team in Toulouse, we show that, for the n-island model of population structure at least, three genomes are enough to separate population size change from structure. The article was submitted this week, and uploaded on the HAL database here.


This figure shows the three IICR plots that can be obtained when three genes are sampled for an n-island model. The figure was obtained for n=10 demes and M=0.1 haploid genomes exchanged per generation.


The figure above shows the spurious changes in population size that would be inferred with the MSMC method for the same demographic model but for three genomes and the three possible sample configurations. In an n-island model there are three ways to sample three genomes: (1) all in the same deme (blue line) , (2) two in the same deme and the third one in another deme (green), (3) all in different demes (black).