Conventional models of gene-sequence evolution assume that all sites
evolve according to a single homogeneous model of evolution or that rates
of evolution vary among sites according to some statistical distribution,
such as the gamma. I describe a likelihood-based model that can
accommodate cases in which different sites in the alignment evolve in
qualitatively different ways, such as might be expected of different
genes in a concatenated alignment or of different regions of the same
gene, such as the stems and loops of ribosomal RNA. We call such
variability 'pattern-heterogeneity'. The method we describe can detect
pattern-heterogeneity without prior partitioning of the data by the
investigator, and simplifies to the gamma rates or homogeneous models as
special cases. It therefore always improves upon these methods and
frequently considerably so. I discuss examples based on an
implementation of the model in a Bayesian MCMC framework.
|