I'll describe our recent efforts to model protein evolution and amino-acid substitutions by (1) designing an easy to use pipeline to estimate amino-acid replacement matrices for specific protein groups, (2) using the protein structure, (3) simplifying mixture models that have been proposed so far. I'll discuss several issues related to models (e.g. complexity trade-off), biology (e.g. independence of nucleotide mutations) and theory (e.g. identifiability, prediction of ancestral sequences).