An implementation of ESM2 in Equinox+JAX
-
Updated
Apr 20, 2026 - Python
An implementation of ESM2 in Equinox+JAX
End-to-end protein function prediction from structure alone: active sites, binding pockets, GO terms, PPI partners, and drug candidates.
Repository of the paper "Exploring sequence landscape of biosynthetic gene clusters with protein language models" published at ICML2024 workshop Machine Learning for Life and Material Science: From Theory to Industry applications
A reproducible in silico pipeline to rationally design and prioritize anti-Listeria antimicrobial peptides. This workflow integrates sequence generation, physicochemical filtering, ESM-2 embedding and multi-criteria scoring to narrow candidates and reduce experimental burden. Open for community evaluation, modification and further development.
Prediction of anti-fungal proteins using protein language models
AI-powered protein mutation analysis platform. Enter any UniProt ID, mutate a residue, and SERAPH, an original ESM2+CNN+BiLSTM model, predicts exactly which secondary structures change and why.
Next-generation protein-protein interaction site (PPIS) prediction integrating ESM-2 protein language model representations with structural & evolutionary features via biophysics-supervised gated fusion and equivariant graph networks.
Antimicrobial peptide discovery via ESM-2 fine-tuning. LoRA training on ESCAPE/GenPept-Curated-2025/PepBenchmark with multilabel evaluation.
Do protein foundation models help predict peptide-HLA class I complex stability on unseen binding grooves? No: a conventional supervised net beats all eleven PLM configurations tested. London AI x Science Hackathon 2026, Serova track.
Similarity search for protein sequences using ESM-2 embeddings and Approximate Nearest Neighbor (ANN) methods.
Experimental-design methods for budgeted protein-epistasis mapping with zero-shot ESM-2 signals; evaluated on GB1.
Functional prediction and evidence-grounded interpretation of KCNH2 missense variants using MaveDB, ESM-2, XGBoost, ClinVar, AlphaMissense and RAG.
Zero-shot variant pathogenicity prediction via spectral covariance analysis of ESM2 hidden states | Claw4S 2026
Residue-residue contact prediction using ESM2 enhanced with template-based structural priors from homologous sequences.
Dual-channel (quantitative + qualitative) predictor for cyclic peptide antimicrobial activity, with annotation learning for unseen cyclization chemistries
Transformer fusion pipeline for protein binding-site prediction using ESM2 protein embeddings, physicochemical surface descriptors, and neighbourhood-aware attention.
Prédiction de la solubilité de protéines recombinantes depuis la séquence (ESM-2 150M fine-tuné + 13 descripteurs biochimiques), identification PDB/UniProt et plan d'expression/purification. Démo Gradio en ligne.
Bioinformatics Kaggle project for prime editing RT activity prediction using leave-one-family-out cross-validation and CLS-oriented modeling.
The protein-engineering design loop as one Rust binary: mutate, fold, QC folders of models into Parquet, ESM-2 variant scoring, a terminal viewer, and a GA4GH TES 1.1 server. Science validated against mdtraj, FreeSASA, cctbx and PLIP in CI.
Unified Python library and CLI for protein structure prediction and inverse folding.
To associate your repository with the esm2 topic, visit your repo's landing page and select "manage topics."