SIB-BLAST

SIB-BLAST refines identification of true and false positives in PSI-BLAST iterative sequence-similarity searches using the SimpleIsBeautiful (SIB) algorithm to improve sequence-based homology detection.

Key Features:

• Iteration comparison: Compares hits from the second PSI-BLAST iteration against those from the final iteration to detect changes across iterations.
• Overlapped-hit metric: Calculates a figure of merit for overlapped hits (present in both iterations) to quantify consistency.
• Re-ranking: Re-ranks PSI-BLAST hits based on the figure of merit to prioritize hits that maintain consistency across iterations.
• Validation against initial profile: Gives precedence to hits validated against the initial profile where model corruption is minimal.
• Model corruption mitigation: Addresses model corruption that occurs in later PSI-BLAST iterations to reduce false positives.
• Use of PSI-BLAST results: Operates on results already generated during a standard PSI-BLAST search, avoiding additional computational burden.
• Sequence-based homology detection: Enhances the accuracy and reliability of sequence-based homology detection in iterative searches.

Scientific Applications:

• Homolog identification: Identification and prioritization of homologous sequences detected by PSI-BLAST iterative searches.
• Distinguishing true/false positives: Improving the discrimination between true and false positives in PSI-BLAST outputs.
• Mitigating iterative profile corruption: Applying to analyses where model corruption in later PSI-BLAST iterations is a concern.

Methodology:

Compare hits from the second and final PSI-BLAST iterations, compute a figure of merit for overlapped hits, re-rank hits by that metric, and validate against the initial profile using results generated during a standard PSI-BLAST search.