Stanisław Dunin-Horkawicz

This paper simulates protein evolution using programmable fitness objectives derived from modern protein models, showing that combined objectives improve evolutionary realism and molecular plausibility.

The paper investigates conformational diversity in poly-HAMP arrays and shows that axial helix rotation may be a conserved mechanism for signal transduction across different HAMP-containing systems.

pLAST improves plasmid classification by combining ORF similarity and gene neighborhood conservation, outperforming existing methods in detecting functional modules and types.

TREAD improves repeat annotation by using protein embeddings instead of sequence profiles, outperforming HMMER in low-data and high-divergence scenarios.

This study identifies Arc/Arg3.1 as a regulator of PABPN1 nuclear speckles through direct binding, impacting mRNA processing during neuronal activity.

This paper evaluates AlphaFold2's ability to model coiled-coil domains, showing improved prediction of oligomeric states compared to previous methods.

The paper introduces pLM-Repeat, a tool using protein language model embeddings to detect repetitive sequences, achieving comparable sensitivity to HHrepID with faster performance.

The paper uses AlphaFold2 to generate 5000 models of the HAMP domain, revealing its conformational variability and structural conservation across families.

pLM-BLAST detects distant homology by comparing sequence embeddings from ProtT5, achieving accuracy comparable to HHsearch while being faster and capable of local alignments.

This paper develops deep learning models to predict and design cofactor specificity in Rossmann-fold proteins, achieving nearly perfect accuracy in benchmark tests.