Thomas Litfin

The paper identifies viruses and proviruses associated with Asgard archaea and reveals virus-mediated interactions with bacteria, expanding understanding of microbial symbiosis in eukaryogenesis.

LMI4Boltz optimizes VRAM usage to predict large macromolecular complexes on consumer-grade hardware without sacrificing accuracy.

SPfast improves protein structure alignment speed and sensitivity, enabling large-scale comparisons and uncovering new biological insights.

This paper introduces an RNA language model based on multiple sequence alignment, showing it can accurately predict RNA structure from sequence data.

This study evaluates deep learning models for RNA base-pairing prediction, showing that SPOT-RNA can accurately predict RNA-RNA interactions in unseen structures and without monomeric structures.

The paper builds a large RNA sequence database (MARS) and integrates it with RNAcmap to improve homology search and RNA structure/function inference.

The paper improves RNA homology detection and alignment using an automated iterative search with an expanded database, achieving performance comparable to Rfam.

The paper introduces SPOT-Contact-Single, a method that improves single-sequence-based protein contact map prediction using a transformer language model, achieving better performance than existing methods on proteins without homologs.

RNAcmap automatically predicts RNA contact maps using evolutionary coupling analysis, outperforming manual methods and showing consistency across RNA sequences.

The paper evaluates 43 methods for predicting intrinsically disordered regions and binding regions in proteins, finding that deep learning methods outperform older physicochemical approaches but binding prediction remains challenging.