Class-defining characteristics in the mouse heavy chains of variable domains

Analysis of residue correlation in over 2700 mouse heavy chains of the V-H domains was carried out on three hierarchical levels. At the 'position' lev el, statistical analysis revealed 45 positions that conserve similar residu es in almost all chains. At the 'fragment' level, the focus of investigatio n shifted to the study of combinations of amino acids in strands and loops. It was found that no more than 10 patterns were sufficient for describing strands and loops in the chains. At the 'sequence' level, we determined all possible combinations of these patterns and classified the mouse heavy cha ins. Comparison of the sequences in the eight classes revealed residues at the class-determining positions that were unique to each class, Because a s trong correlation of residues was found, one only needs several residues to classify a sequence. It follows that no all residue alignment procedure is necessary to divide sequences into classes. An important corollary of our approach is the possibility of predicting residues in an incomplete sequenc e from a small sequence fragment. On the basis of our analysis of mouse hea vy chains we hypothesize about the presently unknown mouse V-H germline rep ertoire.