Regions predicted on HIV sequences by hidden Markov models trained on EIAV sequences. The graphic displays the regions predicted on the HIV-1 HXB2 sequence by our mathematical models (—). The schematic organization of HIV SU with the position of the variable regions (hatched boxes) and of the constant regions (—), as defined by classical amino-acid alignments, is represented under the graphic. (A) HMM of order 5 with 2 hidden states trained on nucleotide sequences of EIAV SU. (B) First order HMM with 2 hidden states trained on deduced amino-acid sequences of EIAV SU.