This page contains information regarding a program which uses a novel criterion and method to classify amino acids. The goal is to identify sets of amino acids with a high probability of change between elements of the set but small probability of change between different sets by using amino acid replacement matrices and their eigenvectors. After identification of the subsets the quality of the partition is assessed with a conductance measure. A detailed description is given in in Kosiol et al. (2004), Journal of Theoretical Biology 228:97-106.

AIS (“Almost Invariant Sets”) is a computer program to implement these ideas, which has been written in ANSI C.

Downloading, Compiling and Running AIS: You can download the AIS program. To execute on a UNIX or linux system:

tar -zxvf ais.tar.gz
make clean
make
ais

Two examples of amino acid replacement matrices, PAM and WAG, are provided with the code. AIS can be used on any amino acid matrix. You will need a file which contains the amino acid rate matrix (often denoted Q), a file including the amino acid equilibrium frequencies (often denoted p) and a file including the right eigenvectors of the amino acid probability matrix (the columns of the matrix U in the notation of Liò and Goldman, 1998, Models of molecular evolution and phylogeny, Genome Research 8:1233-1244). The right eigenvectors should be ordered according to the absolute value of their eigenvalues. Programs like Mathematica will do this automatically for you when calculating the eigenvectors.