{"messages":[{"status":"ok","category":"all"}], "collection":[{"title":"Illuminating phage dark matter beyond the reach of structural alignments with PhageFlare","authors":"Wang, C.; Xu, Y.; Li, Z.; Zhang, L.; Zhang, L.","author_corresponding":"Lu Zhang","author_corresponding_institution":"Hong Kong Baptist University","doi":"10.64898\/2026.10.28.762509","date":"2026-10-29","version":"1","type":"new results","license":"cc_by","category":"microbiology","jatsxml":"https:\/\/www.biorxiv.org\/content\/early\/2026\/10\/29\/2026.10.28.762509.source.xml","abstract":"Many phage proteins remain functionally uncharacterized despite advances in structural-similarity-based annotation. We introduce PhageFlare, a deep learning framework for predicting phage protein functions lacking significant structural matches to annotated references. Across three structure-disjoint benchmarks, where no significant Foldseek match was detected between training and test proteins (E-value thresholds of 10-3, 10-4, and 10-2), PhageFlare achieved macro-averaged F1 scores of 0.871, 0.889, and 0.738 outperforming the strongest baseline (0.789, 0.814, and 0.732). Systematic evaluations demonstrated that protein language model selection critically influences performance, with ESM-C achieving stronger overall performance than ESM-2 and ProtBert. Embedding-guided case studies revealed local structural correspondences among proteins with limited global similarity. Applied to the efam and Metagenomic Gut Virus datasets, PhageFlare annotated 38.0% and 51.8% of proteins unresolved by Foldseek, respectively. Furthermore, PhageFlare identified lysogenic potential in 1,866 prophage-designated genomes misclassified as lytic by BACPHLIP. These results show that PhageFlare complements structural-similarity-based annotation and expands functional discovery across viral dark matter.","published":"NA","server":"bioRxiv"}]}