PSI-Search 2 (EBI)

PSI-Search 2 (EBI) detects distantly related protein sequences by combining SSEARCH (Smith-Waterman local alignment) with iterative profile construction analogous to PSI-BLAST to improve sensitivity and selectivity of homology searches.

Key Features:

• Optimal Alignment Strategy: Uses SSEARCH, the Smith-Waterman local alignment implementation, to perform optimal local alignments and detect subtle sequence similarities.
• Profile Construction with Boundary-Masking: Builds position-specific scoring profiles similar to PSI-BLAST and offers an optional sequence boundary-masking procedure to prevent homologous over-extension (HOE).
• Iterative PSSM Searches: Performs iterative similarity searches using position-specific score matrices (PSSMs) to refine profiles and detect remote homologs.
• Improved Selectivity and Reduced False Positives: Achieves approximately four-fold better selectivity than PSI-BLAST while maintaining similar sensitivity at 50% and 60% family coverage, and produces 2–4 times fewer false positives than JackHMMER while being only marginally (~5%) less sensitive.

Scientific Applications:

• Identification of Distant Homologs: Detection of remotely related protein sequences across diverse datasets and protein families.
• Evolutionary Studies: Comparative analyses and inference of evolutionary relationships among protein families.
• Functional Annotation: Assignment or refinement of protein functional annotations based on detected homology.
• Discovery of Novel Domains: Detection and characterization of previously unrecognized protein domains through sensitive homology searches.

Methodology:

Iterative similarity searches using position-specific score matrices (PSSMs), SSEARCH (Smith-Waterman) local alignments, PSI-BLAST-like profile construction, and optional sequence boundary-masking to prevent homologous over-extension (HOE).