Substrate-Aware Zero-Shot Predictors for Non-Native Enzyme Activities

The `data/` folder is organized as follows: `lib/` contains CSV files with sequence (AAs), fitness, and selectivity (when applicable) data. Each file is named using the format `{enzyme}_{substrate}.csv`. Note: `ParLQ-a` is sometime abbreviated as `ParLQ`. `seq/` contains FASTA files for each enzyme sequence. `structure/` contains the structural files in PDB format. `ParLQ.pdb` is taken from Yang and Lal et al. (2025). `PfTrpB.pdb` corresponds to chain A of PDB entry `5DW0`. `Rma.pdb` corresponds to PDB entry `3CP5`. `apo/` contains apo (substrate-free) structures in PDB format. `clean/` contains cleaned structures (solvent removed) in PDB format. `docked/` contains docked structures in CIF format, generated using AF3 or Chai. See `docked/README.md` for additional information on the docked structures. `evmodel/` contains EVcouplings model files (`.model`) for each enzyme, which are used to calculate EVmutation scores. `evmodel_dets/` contains detailed metadata for the corresponding EVcouplings models.