BLOSUM

BLOSUM, short for Blocks Substitution Matrix, is a family of substitution matrices used to score alignments between protein sequences. Each matrix provides a table of scores for substituting one amino acid for another, expressed as log-odds values. Positive scores indicate substitutions observed more frequently than expected by chance, while negative scores indicate less frequent substitutions. The matrices are used to quantify evolutionary likelihoods in sequence alignment algorithms.

The BLOSUM matrices were introduced in 1992 by Steven Henikoff and Jorja Henikoff. They were derived from local sequence alignments within conserved regions, or blocks, of protein families stored in the BLOCKS database. For each matrix, observed substitution frequencies within these blocks are compared to background amino acid frequencies to produce log-odds scores. A key design feature is that a single matrix is associated with a threshold of sequence identity used to define the blocks.

The number in a BLOSUM name (for example, BLOSUM62, BLOSUM80, BLOSUM45) indicates the minimum percent identity required of blocks used to construct that matrix. Higher numbers reflect more closely related sequences and yield matrices with more positive scores for conservative substitutions, whereas lower numbers are better suited for more divergent sequences. BLOSUM62 is the most widely used default in many protein similarity searches.

In practice, BLOSUM matrices are employed in various sequence alignment tools, including BLAST and FASTA, to compute alignment scores in conjunction with gap penalties. They provide a practical, empirically derived alternative to other substitution models and remain a standard choice for protein homology detection and annotation tasks.