This site contains information about, and applications of pair-to-pair (P2P) amino acids substitution matrices retrieved from high quality protein sequence alignments (Blocks DB). A series of different P2P matrices have been derived (see documentation which include a score for each of the 400*400 possible simultaneous substitutions of pairs of amino acids (40,205 unique combinations considering symmetry).
Currently the scores in the default P2P matrix reflect the likelihood for each substitution of amino acids to be accompanied by another substitution at a position not necessarily close along the sequence, but close in space (i.e. in physical contact). As such, the matrix integrates evolutionary information (derived from high quality multiple sequence alignments) with structural data. There are, however, no conceptual limitations to derive P2P matrices that will reflect other types of correlated substitutions. Correlated mutations may not indeed necessarily reflect a physical contact.
This site allows users to analyze the elements of the P2P substitution matrices, performing online calculation of the correlation between any two sites in the user's multiple sequence alignments using the P2P matrix. It also provides convienient on line graphical environment for analysis and visualization of the results. Finally, the source code for predicting acssociation between sites in multiple sequence alignments using P2P matrices, can be downloaded for different Unix/Linux platforms as well as for Windows.
Send questions and comments to Eran Eyal - e y a l @ c c b b . p i t t . e d u