alterlab-esm

To design proteins with multimodal AI, you provide partial sequences with masked regions and use ESMProtein with GenerationConfig to orchestrate multi-track generation across sequence, structure, and function tracks using ESM3 or ESM C models.

Function-conditioned protein generation creates sequences tailored to specific functional attributes. This Skill leverages ESM3 models to condition outputs on functional data, enabling iterative chain-of-thought workflows for targeted protein design.

Generating protein embeddings for ML tasks involves extracting numerical representations from sequences. This Skill uses local models like esmc-600m or Forge deployments to produce embeddings for clustering, similarity analyses, and downstream machine learning.

Yes, you can use local ESM models for protein structure prediction. This Skill supports local models such as esm3-sm-open-v1 and esmc-600m, and also enables inverse folding to design sequences that fold into target structures.

Yes, protein structure prediction with ESM3 supports scalable batch inference. This Skill integrates with Forge API deployments, allowing you to run scalable inference and batch processing for large-scale protein design workflows.

Yes, you need the Python ESM SDK to run multimodal protein analysis. This Skill requires a Python environment with the ESM SDK installed to orchestrate multi-track generation and manage local models or Forge deployments.