RF2

GitHub repo for RoseTTAFold2

Update May 16, 2024:

• Fixes for training
• New option -cyclize for prediction of cyclic peptides

Update Apr 12, 2024:

• Updated config file
• Better memory efficiency during inference
• Symmetry bugfix

Installation

1. Clone the package

git clone https://github.com/uw-ipd/RoseTTAFold2.git
cd RoseTTAFold2

2. Create conda environment

# create conda environment for RoseTTAFold2
conda env create -f RF2-linux.yml

You also need to install NVIDIA's SE(3)-Transformer (please use SE3Transformer in this repo to install).

conda activate RF2
cd SE3Transformer
pip install --no-cache-dir -r requirements.txt
python setup.py install
cd ..

3. Download pre-trained weights under network directory

cd network
wget https://files.ipd.uw.edu/dimaio/RF2_jan24.tgz
tar xvfz RF2_jan24.tgz
cd ..

4. Download sequence and structure databases

# uniref30 [46G]
wget http://wwwuser.gwdg.de/~compbiol/uniclust/2020_06/UniRef30_2020_06_hhsuite.tar.gz
mkdir -p UniRef30_2020_06
tar xfz UniRef30_2020_06_hhsuite.tar.gz -C ./UniRef30_2020_06

# BFD [272G]
wget https://bfd.mmseqs.com/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt.tar.gz
mkdir -p bfd
tar xfz bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt.tar.gz -C ./bfd

# structure templates (including *_a3m.ffdata, *_a3m.ffindex)
wget https://files.ipd.uw.edu/pub/RoseTTAFold/pdb100_2021Mar03.tar.gz
tar xfz pdb100_2021Mar03.tar.gz

Examples

Prepare to run

conda activate RF2
cd example

Example 1: predicting the structure of a monomer

../run_RF2.sh rcsb_pdb_7UGF.fasta -o 7UGF

Example 2: predicting the structure of a heterodimer with paired MSA

../run_RF2.sh rcsb_pdb_8HBN.fasta --pair -o 8HBN

Example 3: predicting the structure of a heterotrimer with paired MSA

../run_RF2.sh rcsb_pdb_7ZLR.fasta --pair -o 7ZLR

Example 4: predicting the structure of a C6-symmetric homodimer

../run_RF2.sh rcsb_pdb_7YTB.fasta --symm C6 -o 7YTB

Example 5: predicting the structure of a C3-symmetric heterodimer (A₃B₃ complex) with paired MSA

../run_RF2.sh rcsb_pdb_7LAW.fasta --symm C3 --pair -o 7LAW

Expected outputs

Predictions will be output to the folder 1XXX/models/model_final.pdb. B-factors show the predicted LDDT. A json file and .npz file give additional accuracy information.

Additional information

The script run_RF2.sh has a few extra options that may be useful for runs:

Usage: run_RF2.sh [-o|--outdir name] [-s|--symm symmgroup] [-p|--pair] [-h|--hhpred] input1.fasta ... inputN.fasta
Options:
     -o|--outdir name: Write to this output directory
     -s|--symm symmgroup (BETA): run with the specified spacegroup.
                              Understands Cn, Dn, T, I, O (with n an integer).
     -p|--pair: If more than one chain is provided, pair MSAs based on taxonomy ID.
     -h|--hhpred: Run hhpred to generate templates