| Popis: |
Microbiome research has opened new frontiers in public health and environmental stewardship. However, protein sequence database for these complex microbial communities are often incomplete or unavailable, which limits options for spectral annotation. Spectral library search is an efficient method for MS/MS identification, but library sizes can be prohibitively large for microbiome research. Standard techniques apply a precursor ion window to filter candidates for an exact match, which can easily overlook many possible homologous matches. Emerging open library search methods appear promising, but have yet to be tested at the scale necessary for microbial communities. This calls for an efficient open spectral library search approach which can perform open search across a spectral library within a reasonable time frame. As a solution we present a GPU-accelerated, highly efficient pairwise similarity algorithm which can shortlist candidate spectra from a spectral library after performing all to all comparison. Our preliminary results show that open search for 35,000 spectra against a library of 1.18 million spectra takes approximately 45 mins, which is similar to a database search for a small bacterial organism. |
