Two sources of discrepancies between structure-based and profile-based position similarities: misalignment and structural or functional specificity. A. Manual alignment of fragments of human glyoxalase II (PDB ID 1qh5A) and bacterial penicillinase (PDB ID 1sml). Strands b are shifted by one register, so that structurally similar residues correspond to different sequence positions. In these strands, sidechains are shown for the residues involved in Zn binding (D134 and S 185, respectively). (B-D) Alignments produced for these fragments by different methods, with corresponding positional P-values. Secondary structure elements (beta stands a and b) are shown with arrows. Structurally equivalent residues are connected with lines, residues involved in Zn binding are boxed. B. DALI correctly reproduces structural alignment of strands a, resulting in relatively high positional P-values in this region. Alignment of strands b is also structurally correct, but their shift in the two structures results in discrepancy between structural alignment and profile content, as reflected by low P-values. C. MAMMOTH misaligns strands a by one register. Low P-values indicate significant profile dissimilarity between aligned positions in both regions a and b. D. Profile-based alignment is in good accord with profile content, as shown by higher P-values for regions a and b. Note that alignment in region b differs from the structure-based alignment shown in B.