PoSSuMsearch2

PoSSuMsearch2 accelerates database searches for protein family classification by using position-specific scoring matrices family models (PSSM-FMs) to prefilter sequences prior to profile hidden Markov model (pHMM) searches, reducing computational cost while retaining pHMM sensitivity.

Key Features:

• Search Space Reduction: Employs PSSM-FMs to prefilter sequences, reducing candidates for pHMM-based searches and retaining over 99.5% of original findings.
• Speed and Efficiency: Combines full-text indexing, exact p-value computation of PSSM match scores, and fast fragment chaining to achieve more than a 64-fold speedup compared to hmmsearch.
• High Performance: In experiments, reduced sequences needing pHMM analysis to 0.80% of all sequences, yielding an overall speedup of 43 over unfiltered searches and a lossless filtering speedup of 92 for hmmsearch on UniProtKB/Swiss-Prop.

Scientific Applications:

• Large-scale genomic analyses: Enables rapid and accurate protein family classification in large-scale genomic datasets.
• Genome annotation pipelines: Prefilters sequence candidates to reduce the computational cost of pHMM-based genome annotation.

Methodology:

Prefilters sequences using PSSM-FMs; uses full-text indexing, exact p-value computation of PSSM match scores, and fast fragment chaining; then performs subsequent pHMM-based searches (e.g., hmmsearch) on the reduced candidate set.