Burkhard Rost

The paper explores using protein embeddings for nearest neighbor search to detect remote homology, highlighting challenges with multi-domain proteins and proposing solutions.

EMBER3D predicts protein 3D structures from single sequences rapidly, enabling real-time analysis of mutation effects and structural changes.

The paper introduces GRASP, a method for ancestral sequence reconstruction that handles large datasets and indels, demonstrating its ability to generate biologically active enzyme ancestors.

SETH predicts residue disorder from protein embeddings using ProtT5, outperforming complex methods while being faster and capturing nuanced disorder variations.

The paper introduces bindAdjust, a method that improves binding predictions by incorporating structural information from AlphaFold2, enhancing recall for small molecule binding and expanding predictions to larger binding sites.

TMbed predicts transmembrane regions using protein language model embeddings, achieving high accuracy and low false positives for proteome-wide analysis.

This paper presents a method using protein language model embeddings for alignment-free protein structure prediction, achieving performance comparable to co-evolution-based methods at lower computational cost.

This paper introduces ProtTucker, a contrastive learning method using protein embeddings from pLMs to improve recognition of distant homologous relationships in the midnight zone of protein similarity.

The paper introduces bindEmbed21, a deep learning method using protein embeddings to predict binding residues for metal ions, nucleic acids, and small molecules, achieving high accuracy without using MSAs or structural data.

This paper trains and evaluates language models on protein sequences to predict structural and localization features, achieving state-of-the-art results without evolutionary information.

The paper improves the purity of protein functional families by clustering them using sequence embeddings, doubling the number of pure clusters per family.

PredictProtein provides protein sequence analysis tools, offering structure and function predictions with updates including deep learning and DNA/RNA/protein binding predictions.

This paper proposes using deep learning embeddings to transfer GO annotations based on protein sequence proximity rather than homology, achieving performance comparable to existing methods.

The paper introduces a lightweight deep learning method using protein language model embeddings to predict protein subcellular localization, outperforming state-of-the-art methods without using multiple sequence alignments.

The paper generates 872 structural models of SARS-CoV-2 proteins and creates a structural coverage map to help researchers understand viral mechanisms and immune disruption.

This paper develops an evolutionary model of protein secondary structure to reveal biological relationships not evident in amino acid-based analysis.

The paper evaluates variant effect prediction methods against DMS data, finding that some methods capture aspects of variant effects but none perfectly, with Envision performing best for precise effect prediction.

This paper introduces an end-to-end multitask learning approach using language models to predict protein structure and function without evolutionary alignments, achieving faster predictions with moderate accuracy.

The paper analyzes the impact of common and rare single amino acid variants in helical transmembrane proteins, finding common variants have stronger effects and are enriched in membrane helices, while 7TM proteins are depleted of SAVs.

This paper introduces SeqVec, a method to represent protein sequences as continuous vectors using deep learning, achieving better prediction results than methods relying on evolutionary information.

The paper compares predicted protein localization spectra across ten eukaryotic species to infer evolutionary relationships.

The paper introduces funtrp, a method to classify protein positions as neutral, rheostat, or toggle based on mutational impact, showing these types correlate with functional effects and improve variant effect prediction.

The paper predicts novel protein-protein interactions for eight model organisms using sequence-based methods and provides these predictions through a web interface and downloadable data.

The paper introduces a computational method to identify type III effectors in bacteria, finding that recognition signals span the entire protein sequence and identifying new effectors across prokaryotic genomes.

The paper demonstrates that linked annotations from text corpora can complement database annotations by showing novel subcellular localization findings in a curated corpus.