Back

Scaling and democratising structure-based protein function prediction with metagenomic-deepFRI

Bezshapkin, V.; Schymik, F.; Kucharski, P.; Szczerbiak, P.; Wojciechowski, J. W.; SZYDLOWSKI, L. M.; Kosciolek, T.

2026-04-29 bioinformatics
10.64898/2026.04.27.720983 bioRxiv
Show abstract

High-quality protein structure models have become widely available, offering insights into protein function, yet they remain underutilized. Here, we introduce metagenomic-deepFRI, a framework incorporating structural templates into functional annotation pipelines at speeds comparable to sequence-alignment methods. Notably, structural features improved GO term prediction confidence and Information Content by up to 50%. Applied to metagenomic datasets, the framework achieves nearly 90% annotation coverage, enabling protein function inference without explicit orthology-based transfer.

Matching journals

The top 2 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.