A pipeline for evaluating protein structure quality using AF3Score.
conda create -n af3score python=3.11 conda activate af3score conda install gxx_linux-64 gxx_impl_linux-64 gcc_linux-64 gcc_impl_linux-64=13.2.0
git clone https://github.com/Mingchenchen/AF3Score.git cd AF3Score/ # Install Python dependencies pip install -r dev-requirements.txt pip install --no-deps -e . build_data # Install additional dependencies conda install -c conda-forge biopython h5py pandas
Download Databases:
bash fetch_databases.sh <DB_DIR> # Replace <DB_DIR> with your database directory
Install HMMER:
mkdir ~/hmmer_build ~/hmmer wget http://eddylab.org/software/hmmer/hmmer-3.4.tar.gz -P ~/hmmer_build cd ~/hmmer_build tar -zxf hmmer-3.4.tar.gz cd hmmer-3.4 ./configure --prefix=~/hmmer make -j8 make install
Add HMMER to your PATH:
export PATH="~/hmmer/bin:$PATH"
Verify installation:
hmmsearch -h
The AF3Score pipeline is designed for high-throughput evaluation of protein structures. It consists of two primary scripts tailored for single-batch or multi-batch processing on high-performance computing (HPC) clusters.
AF3score_pipeline.sh is the core utility used to process a single directory of PDB files.
Usage:
Before running the pipeline on a shell cluster, you must configure the variables within AF3score_pipeline.sh.
| Variable | Description | Example Value |
|---|---|---|
PYTHON_EXEC |
Path to the specific Conda environment Python binary. | ~/anaconda3/envs/af3score/bin/python |
slurm_partition |
Target GPU partitions for job submission. | gpu1,gpu2 |
slurm_nodelist |
Specific nodes assigned for the computation. | c06b14n[05-06],c06b19n[05-06] |
Run the pipeline:
./AF3score_pipeline.sh <input_pdb_dir> <output_dir> <num_jobs>
<input_pdb_dir>: Path to the directory containing your input.pdbfiles.<output_dir>: Target directory where AF3Score metrics and results will be saved.<num_jobs>: The number of parallel jobs to launch.
For users handling multiple datasets across several directories, use the multi-directory wrapper AF3score_mutildir.sh.
The pipeline generates the following scoring metrics:
| Metric | Level | Description |
|---|---|---|
| pTM | Global / Per-chain | Predicted TM-score: Measures the overall topological accuracy of the global structure. |
| ipTM | Global / Inter-chain | Interface pTM: Assesses the accuracy of the interfaces between different protein chains. |
| pLDDT | Per-residue / Per-chain | Predicted Local Distance Difference Test: A per-residue confidence score (0-100). Higher values indicate higher local structure stability. |
| PAE | Per-chain | Predicted Aligned Error: The expected distance error (in Å) between pairs of residues. Lower values indicate higher confidence in relative positioning. |
| ipSAE | Inter-chain | interaction prediction Score from Aligned Errors: Specifically focuses on the binding interface of two chains. |
Global level metrics are evaluates the quality of the overall structure. Per-chain metrics are focused on the quality of individual chains. Inter-chain metrics are designed to assess the quality of the docking between two chains.
For more information about AlphaFold3, please visit their GitHub Repository