RareFold

All compressions are made with zstd here. You can get it here (and it's amazing - thank me later): https://github.com/facebook/zstd

params50000.npy - parameters for RareFold single-chain protein structure prediction and design

finetuned_params25000.npy - parameters for RareFold peptide binder design (less tuned)

1ssc.tar.zst - design results for PDB ID 1ssc

val.tar.zst - validation results

test.tar.zst - RareFold test set predictions using the baseline model (The same data AF3 is/can be evaluated on)

test_LC.tar.zst - RareFold test set predictions using the large-crop fine-tuned weights

af3.tar.zst - AF3 predictions

native_valid_structures.tar.zst - native validation structures from the PDB

train_meta_20.csv - meta for train/val/test splits and 20% sequence identity cluster annotations

aa_count_per_chain.csv - Number of AAs of each type per chain id

aa_hist.csv - total occurrence of AA types in the dataset

immunogenicity_analysis.xlsx - results from the immunogenicity analysis of the peptide binders compared to the wild-type binder