esm-protein-language-model

To design a novel protein sequence, you can use the ESM3 and ESM C protein language models to generate candidate amino acid sequences tailored to desired structural or functional properties for de novo protein design.

Yes, you can predict protein 3D atomic structure from an amino acid sequence. The ESM protein language models infer atomic-level 3D coordinates directly from the provided sequence inputs.

You need either a local GPU or the EvolutionaryScale Forge cloud API to run protein language model inference. These computational resources are required to execute the ESM3 and ESM C model operations.

Inverse folding in protein design is the process of designing amino acid sequences that will fold into a specific target 3D structure. This Skill uses ESM models to generate sequences matching your structural constraints.

To generate protein embeddings for machine learning, the Skill uses ESM models to create fixed-length vector representations of protein sequences. These embeddings facilitate downstream tasks like classification and clustering.

The best way to cluster similar designed protein structures is to generate fixed-length protein embeddings using the ESM models. You can then apply standard machine learning clustering algorithms to these vector representations.