SimpleFold: Folding Proteins Is Simpler Than You Think

Summary

SimpleFold: Folding Proteins is Simpler than You Think Introduction We introduce SimpleFold, the first flow-matching based protein folding model that solely uses general purpose transformer layers. SimpleFold does not rely on expensive modules like triangle attention or pair representation biases, and is trained via a generative flow-matching objective. We scale SimpleFold to 3B parameters and train it on more than 8.6M distilled protein structures together with experimental PDB data. To the best of our knowledge, SimpleFold is the largest scale folding model ever developed. On standard folding benchmarks, SimpleFold-3B model achieves competitive performance compared to state-of-the-art baselines. Due to its generative training objective, SimpleFold also demonstrates strong performance in ensemble prediction. SimpleFold challenges the reliance on complex domain-specific architectures designs in folding, highlighting an alternative yet important avenue of progress in protein structure prediction. Installation To install simplefold package from github repository, run git clone https://github.com/apple/ml-simplefold.git cd ml-simplefold python -m pip install -U pip build; pip install -e . pip install git+https://github.com/facebookresearch/esm.git # Optional for MLX backend Example We provide a jupyter notebook sample.ipynb to predict protein structures from example protein sequences. Inference Once you have simplefold package installed, you can predict the protein structure from target fasta file(s) via the following command line. We provide support for both PyTorch and MLX (recommended for Apple hardware) backends in inference. simplefold \ --simplefold_model simplefold_100M \ # specify folding model in simplefold_100M/360M/700M/1.1B/1.6B/3B --num_steps 500 --tau 0.01 \ # specify inference setting --nsample_per_protein 1 \ # number of generated conformers per target --plddt \ # output pLDDT --fasta_path [FASTA_PATH] \ # path to the target fasta directory or file --ou...