small-protein-gen

A collection of generative models on small proteins (<=128) trained on my macbook pro.

Models

Models are labelled according to their reference's name, but implementations (and hence performance) probably differ with the original.

Name	Description	Reference
foldingdiff	Diffuse backbone torsion and bond angles in $\mathbb{T}^{6\times (N - 1)}$.	Wu et al. (2024)

Data

Dataset used is a non-redundant version of the CATH dataset with max 40% sequence similarity. Chains are filtered to be within 40-128 residues in length (no random crops). This results in a total of 15,696 monomers.

Feedback

Feedback is welcome! Raise an issue or drop a PR.