esm

Integrate ESM-C embeddings into molecular dynamics simulations. Generate per-residue embeddings, map them to structures, and export for OpenMM or GROMACS restraint weights or collective variables in MD engines.

Discover how Biohub Platform safety filters detect controlled pathogen sequences using curated blacklists of high-risk entries and regulated keywords to flag potential concerns automatically.

Master fine-tuning ESMC models on custom protein datasets. This guide details loading checkpoints, tokenizing sequences, and optimizing loss for advanced protein analysis.

Discover how VQVAE codebook training transforms protein structures into discrete tokens for language models. Learn to represent complex biological data efficiently.

Learn to use the Forge API retry mechanism for reliable batch inference. Process thousands of sequences automatically adapting to API rate limits with tenacity and AIMD rate limiting.

Learn to compute and visualize predicted aligned error PAE matrices for protein structure. Understand inter-residue distance confidence using novel methods from the Biohub ESM repository.

Discover how ESMFold2 predicts mixed protein-DNA-RNA complexes. Learn about its type-aware input pipeline for unified token representations and end-to-end structure prediction of heterogeneous assemblies.

Learn to interpret pLDDT confidence scores and predicted error in ESM Fold. Understand residue uncertainty and model accuracy from Biohub/esm for precise protein structure analysis.

Learn how the function token decoder maps InterPro annotations to protein sequences using a compact Transformer stack for efficient per-position classification and contiguous annotation generation.

Easily load ESMProtein from PDB files using ESMProtein.from_pdb() to get sequence coordinates and confidence scores in one line.

Easily configure LogitsConfig to return embeddings for downstream ML tasks by setting sequence and return_embeddings to True in the Biohub esm client. Logits() method extracts dense per-residue embeddings.

Design protein binders with specific pocket contact requirements using Biohub/esm's guided generation and PocketConditioning constraints. Learn this derivative-free decoding engine.