RoseTTAFold2-PPI

A fast deep learning method for large-scale protein-protein interaction screening.

Installation

1. Download the environment image using one of the following commands:

   • wget --no-check-certificate https://conglab.swmed.edu/humanPPI/SE3nv.sif
   • wget --no-check-certificate http://prodata.swmed.edu/humanPPI/bulk_download/SE3nv.sif

2. Clone the repository:

   git clone https://github.com/CongLabCode/RoseTTAFold2-PPI.git
   

3. Download the weights to RoseTTAFold2-PPI/src/model:

   cd RoseTTAFold2-PPI/src/models
   wget --no-check-certificate https://conglab.swmed.edu/humanPPI/downloads/RF2-PPI.pt 
   

Usage

To run RoseTTAFold2-PPI using the Singularity image, use the following command:

singularity exec \
  --bind /path/to/input_and_output_directory:/work/users \
  --bind /path/to/rosettafold2-ppi/directory:/home/RoseTTAFold2-PPI \
  --nv SE3nv.sif \
  /bin/bash -c "cd /work/users && python /home/RoseTTAFold2-PPI/src/predict_list_PPI.py input_file"

Input File Format

For the input_file, e.g., examples/input_file, each line should contain two columns:

1. File path of the concatenated pairwise multiple sequence alignment (MSA) input.
2. Length of the first protein.

Note: When using Singularity, paths should be relative to the directories mounted inside the container. If you prefer to use absolute paths, ensure they reference the file paths inside the container after mounting the directories.

Output File

The output file will be saved as [input_filename].npz, where input_filename is the name of your input file.

Test

cd RoseTTAFold2-PPI
exec_dir=$(pwd)
singularity exec \
    --bind $exec_dir:/home/RoseTTAFold2-PPI \
    --nv SE3nv.sif \
    /bin/bash -c "cd /home/RoseTTAFold2-PPI && python /home/RoseTTAFold2-PPI/src/predict_list_PPI.py examples/test.list"

The command will generate test.list.log and test.list.npz under RoseTTAFold2-PPI/examples which should be the same as files under examples/expected_output.

Note: The performance is affected by the quality of the multiple sequence alignment. Our benchmarks suggest that trimming low-quality regions, such as poorly conserved intrinsically disordered regions, enhances the accuracy of RoseTTAFold2-PPI.

.

Reference

Jing Zhang*, Ian R Humphreys*, Jimin Pei*, Jinuk Kim, Chulwon Choi, Rongqing Yuan, Jesse Durham, Siqi Liu, Hee-Jung Choi, Minkyung Baek, David Baker, Qian Cong. Computing the Human Interactome. (https://www.biorxiv.org/content/10.1101/2024.10.01.615885v1)