Improving real-time search through deep learning-enhanced fragment ion indexing

Masking fragment ion index entries using Carafe in Comet sharply reduces peak search memory use while keeping peptide-spectrum match counts comparable across tested search spaces and index settings.

• Why it matters: Fragment ion indexed database searching can require tens of gigabytes of memory, limiting searchable database size, variable modifications, and concurrent real-time searches on instrument acquisition computers.
• What they did: The authors built an index-pruning pipeline in Comet that uses Carafe to predict fragment ion intensities from peptide sequences, writing only predicted observable fragment ion positions into the index.
• The result: Across two search spaces and three index configurations, masked searches used fewer index entries, ran faster than unmasked searches, and found a comparable number of peptide-spectrum matches with no per-spectrum search overhead.

AI summary of the abstract · may contain errors · not medical advice · read the paper.