Nugkta/bgcbench-bigcarp

BiGCARP (BGC-Bench mirror)

This repository redistributes the BiGCARP weights released by the original authors, so that BGC-Bench can download them at a fixed tag. It is not an official release of the original authors, and the weights are not retrained. Licence: MIT, as upstream. Please cite the original paper (below). Intended use: frozen feature extraction in BGC-Bench; see the original paper for the model's training data and limitations.

Upstream

• Paper: Rios-Martinez, Bhattacharya, Amini, Crawford & Yang (2023). Deep self-supervised learning for biosynthetic gene cluster detection and product classification. PLOS Computational Biology 19(5): e1011162. https://doi.org/10.1371/journal.pcbi.1011162
• Code: https://github.com/microsoft/bigcarp
• Weights: Zenodo record 6857704 (v1.0, MIT), https://doi.org/10.5281/zenodo.6857704: bigcarp_esm1bfinetune.pt and final_domain_vocab.json

Format

model.safetensors holds the checkpoint's model_state_dict with unchanged tensor values; vocab.json is final_domain_vocab.json byte for byte; config.json records the ByteNet hyper-parameters.

Model

ByteNet masked language model (32 dilated CNN layers, d_model 256) over Pfam domain names. BGC-Bench uses the final-layer states: their mean over the BGC's domains (256-d), and the state at an A-domain's AMP-binding token.

Files

file	bytes
LICENSE	1,141
config.json	669
model.safetensors	136,173,812
vocab.json	532,383

Use

git clone https://github.com/Nugkta/BGC-Bench && cd BGC-Bench
uv sync
uv run bgcbench run --encoder bigcarp

Citation

@article{bigcarp,
  title   = {Deep Self-Supervised Learning for Biosynthetic Gene Cluster Detection and Product Classification},
  author  = {Rios-Martinez, Carolina and Bhattacharya, Nicholas and Amini, Ava P. and Crawford, Lorin and Yang, Kevin K.},
  journal = {PLOS Computational Biology}, volume = {19}, number = {5}, pages = {e1011162}, year = {2023},
  doi     = {10.1371/journal.pcbi.1011162}
}