Résumé
Incomplete lineage sorting (ILS), the interaction between coalescence and
speciation, can generate incongruence between gene trees and species
trees, as can gene duplication (D), transfer (T) and loss (L). These
processes are usually modelled independently, but in reality, ILS can
affect gene copy number polymorphism, i.e., interfere with DTL. This has
been previously recognised, but not treated in a satisfactory way, mainly
because DTL events are naturally modelled forward-in-time, while ILS is
naturally modelled backwards-in-time with the coalescent. Here we consider
the joint action of ILS and DTL on the gene tree/species tree problem in
all its complexity. In particular, we show that the interaction between
ILS and duplications/transfers (without losses) can result in patterns
usually interpreted as resulting from gene loss, and that the realised
rate of D, T and L becomes non-homogeneous in time when ILS is taken into
account. We introduce algorithmic solutions to these problems. Our new
model, the multilocus multispecies coalescent (MLMSC),
which also accounts for any level of linkage between loci, generalises the
multispecies coalescent model and offers a versatile, powerful framework
for proper simulation and inference of gene family evolution.