The abstract is read at the publisher; the summary is JClub's.
Infernal-GPU matches Infernal’s covariance-model search results while enabling longer E-value calibration, revealing Rfam E-values are off by more than twofold for one model in eight.
- Why it matters: Infernal’s approximations—short random-sequence E-value calibration and sequence-only filtering—could reduce accuracy, but the magnitude and where errors occur were largely unknown.
- What they did: The GPU implementation parallelizes the CM scoring self-loop via two prefix scans, then calibrates all 4,227 Rfam models overnight using ten times more random sequence.
- The result: Longer calibrations show E-values for Rfam are off by more than twofold for one model in eight, and exhaustive reference indicates filters discard remote homologs where similarity is mostly base pairing, which Infernal-GPU recovers with one third of the CPU work.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.