Exploring zero-shot structure-based protein fitness prediction

This repository contains data used in Exploring zero-shot structure-based protein fitnessprediction. Directions to use this data can be found on our GitHub repository. experimental_struct_artifacts contains the experimentally determined structures for ProteinGym assays used in our analysis along with the reference file needed to generate ESM inverse folding predictions for these structures in ProteinGym. results contains the prediction results obtained by running SSEmb on the 216 ProteinGym assays being considered in this study. test.tar.gz contains all the structures from ProteinGym as well as MSAs generated using mmseqs2. To use this directory: Setup SSEmb as directed in its repository Download this file and extract it in the data folder.