Detection of structurally aligned protein positions that correspond to dissimilar profile positions. (A, B) Histograms of P-values for FSSP alignments between domains of different sequence identity (log-log scale): A. Sequence identity 15 ± 1%. B. Sequence identity 25 ± 1%. (C, D) Difference of residue frequencies at alignment positions with lowest P-values and the whole population: C. Sequence identity 15 ± 1%, P < 10-14 . D. Sequence identity 25 ± 1%, P < 10-18.