alphaflow

AlphaFlow Example

In order to run AlphaFlow with SBGrid you must:

1. Obtain the alphaflow model weights

2. Obtain MSAs for your sequences

These steps are included in the example script below. We use a test input csv file provided with AlphaFlow.

Edit the project_dir variable to point to your own project directory.

## AlphaFlow Example
## https://github.com/bjing2016/alphaflow
## 
## help@sbgrid.org
## Dec 3, 2024


source /programs/sbgrid.shrc
export ALPHAFLOW_X=20241203_86d7606

verdir=/programs//x86_64-linux//alphaflow/20241203_86d7606
project_dir=/build/source/alphaflow/20241203_86d7606

cd ${project_dir}

## Get weights
## AlphaFlow-PDB distilled
mkdir weights
wget -P ./weights https://storage.googleapis.com/alphaflow/params/alphaflow_pdb_distilled_202402.pt

## Sample input file 
## edit to contain just one sequence for testing
cp ${verdir}/alphaflow/splits/atlas_test.csv ./

## Generate MSAs from colabfold
python.alphaflow  ${verdir}/alphaflow/scripts/mmseqs_query.py --split ./atlas_test.csv --outdir ./msa

## Run model
python.alphaflow ${verdir}/alphaflow/predict.py --mode alphafold --input_csv ${project_dir}/atlas_test.csv  --msa_dir ${project_dir}/msa --weights ${project_dir}/weights/alphaflow_pdb_distilled_202402.pt --samples 1 --outpdb ${project_dir}/out_pdb