A maximum-likelihood approach to analyzing nonoverlapping and overlapping reading frames.

A model is presented for sequence evolution on the basis of which one can analyze combinations of noncoding, singly coding, and multiply coding regions of aligned homologous DNA sequences. It is a generalization of Kimura's (J. Mol. Evol. 16:111-120, 1980) and Li et al.'s (J. Mol. Evol. 36:96-99, 1985) transition-transversion models with selection on replacement substitutions. Based on a hierarchy of hypotheses, one will be able to estimate selection factors and transition and transversion distances for different combinations of regions ranging from many regions, each with their private set of parameters, to one set of parameters for all regions. The method is demonstrated on two aligned HIV1 retroviruses.