BLOSUM

BLOSUM matrices were constructed by Henikoff and Henikoff.

• They derived the BLOSUM matrices from a set of ungapped regions from protein database called the BLOCKS database. The procedure is as follows:
• Whenever two sequence were found to have sequence identity greater than some L%, they were put into the same CLuster. This way they clustered sequences from each BLOCK.
• Then the frequency Aab, of observing a in one cluster aligned against b in another cluster. Correction for the sizes of the clusters is done byweighting each occurence by 1/(n1n2), where n1 and n2 are the respective cluster sizes.
• From Aab they estimated qa and Pab. From these they derived the score matrix entries using the standard equation s(a,b) = log Pab/qa qb.
• The resulting log odds matrices were scaled and rounded to the nearest integer value.

The Matrices for L=62 and L=50 are widely used for pairwise alignment and Database searching.

BLOSUM 62 is the standard for ungapped alignments and BLOSUM50 is the standard for gapped alignments.

(Lower L values correspond to longer evolutionary time and are applicable for more distant searches.)