Janani Durairaj

Stoic: Fast and accurate protein stoichiometry prediction

Stoic predicts protein complex stoichiometry using protein language model embeddings and graph neural networks, achieving fast and accurate results for homomeric and heteromeric complexes.

Capabilities, specificity gaps and training-data dependence of AlphaFold3 across diverse application areas

This paper evaluates AlphaFold3's performance across diverse structural biology tasks, finding uneven accuracy and strong dependence on training data.

Have protein-ligand co-folding methods moved beyond memorisation?

This study evaluates four protein-ligand cofolding methods using a new benchmark dataset, finding they primarily memorize ligand poses rather than predict them de novo.

Reading TEA leaves for de novo protein design

This paper introduces a method for fast de novo protein design using a structure-inspired alphabet and Metropolis sampling to generate sequences without known homologues.

Rewriting protein alphabets with language models

This paper introduces a new 20-letter alphabet, TEA, derived from protein language models, enabling efficient homology searches comparable to structure-based methods without structural data.

The Viral AlphaFold Database of monomers and homodimers reveals conserved protein folds in viruses of bacteria, archaea, and eukaryotes

The paper predicts structures for 27,000 viral proteins using AlphaFold2, revealing conserved folds across viruses infecting bacteria, archaea, and eukaryotes and identifying a new antiviral toxin-antitoxin system.

PLINDER: The protein-ligand interactions dataset and evaluation resource

PLINDER provides a large, annotated PLI dataset with diverse annotations and improved evaluation splits to better assess computational methods for drug design.

Embedding-based alignment: combining protein language models and alignment approaches to detect structural similarities in the twilight-zone

This paper introduces an embedding-based alignment method using protein language models to detect structural similarities in the twilight zone, outperforming existing approaches without training.

Fast and adaptive protein structure representations for machine learning

The paper introduces a fast structure aligner using rotation-invariant shape-mers for protein structures, enabling efficient machine learning analysis of functional and structural similarities.

Integrating structure-based machine learning and co-evolution to investigate specificity in plant sesquiterpene synthases

The paper predicts precursor cation specificity in plant sesquiterpene synthases using machine learning and co-evolution analysis, identifying key residues for enzyme function.

Geometricus Represents Protein Structures as Shape-mers Derived from Moment Invariants

Geometricus embeds protein structures into fixed-length vectors using shape-mers derived from 3D moment invariants, enabling tasks like structure similarity search and clustering.

An Analysis of Characterized Plant Sesquiterpene Synthases

The paper analyzes plant sesquiterpene synthases to identify motifs and unexplored sequence regions, and demonstrates a database application for guiding mutagenesis studies on product specificity.