Mean and standard deviation (in parentheses) of difference in log-likelihood between group-specific and alternative rate matrices (<i>R_Ti</i>) as applied to the validation datasets (<i>V</i>_i) of the cross validation procedure.

R0 is the rate matrix estimate from a complete set of real amino acid sequences. RTi is a rate matrix estimated from a group-specific training dataset. The cross validation procedure was based on 10 replicates of 50% cross validation.