4: Scoring Matrices: PAM, BLOSUM & Log-Odds Mathematics
Chapter 4: Scoring Matrices: PAM, BLOSUM & Log-Odds Mathematics
Nucleotide alignments often treat all mismatches equally. However, in protein coding sequences, amino acids possess vastly different physicochemical properties (size, charge, hydrophobicity, aromaticity). Replacing a hydrophobic Leucine with an Isoleucine rarely disrupts tertiary structure, whereas replacing a hydrophobic Leucine with a charged Arginine or a structural Proline can completely destabilize an alpha-helix. Amino acid scoring matrices quantify these evolutionary replacement probabilities.
Every cell \(S_{ij}\) in a substitution matrix represents the logarithm of the odds that amino acids \(i\) and \(j\) are observed aligned due to shared evolutionary ancestry (homology) versus random chance:
Sij = λ × log2 [ qij / (pi × pj) ]
Where \(q_{ij}\) is the observed target frequency of alignment between amino acids \(i\) and \(j\), \(p_i\) and \(p_j\) are their independent background background frequencies, and \(\lambda\) is a scaling factor (typically half-bits).
1. PAM vs. BLOSUM: Two Opposing Methodologies
- PAM (Point Accepted Mutation): Developed by Margaret Dayhoff in 1978. Derived from global alignments of closely related proteins sharing ≥85% sequence identity (PAM1 = 1 substitution per 100 amino acids). Higher PAM matrices (e.g. PAM250) are calculated mathematically by taking powers of the PAM1 transition probability matrix (\(M^{250}\)) under an explicit Markovian evolutionary model.
- BLOSUM (BLOcks SUbstitution Matrix): Developed by Henikoff and Henikoff in 1992. Directly derived from observed local ungapped alignments ("blocks") in the BLOCKS database. Sequences sharing more than a threshold identity are clustered to reduce redundancy. BLOSUM62 clustered sequences at ≥62% identity; BLOSUM80 clustered at ≥80% identity.
2. Practical Selection Guidelines
# Evolutionary Distance and Matrix Guide:
# -------------------------------------------------------------------------------------
# Target Relationship Sequence Identity Recommended Matrix Application
# -------------------------------------------------------------------------------------
# Very closely related > 75% identity BLOSUM80 / PAM120 Recent paralogs
# Moderately related 40% - 70% identity BLOSUM62 (Default) Standard BLASTp
# Distantly related 20% - 35% identity BLOSUM45 / PAM250 Deep homologs
# -------------------------------------------------------------------------------------Common Pitfalls & Troubleshooting Checklist
- Java Heap Exhaustion (GATK / Picard): If processes terminate with
java.lang.OutOfMemoryError, set your environment variable explicitly:export _JAVA_OPTIONS="-Xmx16g -Djava.io.tmpdir=/tmp". - Index Timestamp Mismatch: If tools complain that a BAM index is missing or out of date, regenerate with
samtools index -@ 4 sample.bam. - Conda Channel Conflicts: Avoid ABI incompatibilities between bioconda and conda-forge by setting strict priority:
conda config --set channel_priority strict. - Chromosome Naming Contamination: Always verify whether reference FASTA headers use
chr1or1; divergent naming causes silent alignment drops.