Complex-prediction

This will be a snakemake pipeline to predict complexes with Combfold.

Usage

This snakemake 🐍 project is powered by pixi 🚀, it handles all of the dependencies (@me if it don't!). Installing the pipeline involves the following steps:

# Install pixi
curl -fsSL https://pixi.sh/install.sh | bash

# Print the help
pixi install

The hardware requirement is a CUDA 11.8 or higher enabled graphics driver. If this is met, analysis can begin!

# Run the whole analysis pipeline from begining to end for all files in `data`
pixi run make

Usage in a HPC context is done with the

pixi run slurm

command, this will launch the Snakemake pipeline with the slurm executor. This means each job is launched as a seperate job with one GPU allocator. This behaviour is supported on the raven HPC system.

Steps in the pipeline

The input data in the form of combfold formatted JSON files from data are copied over to the results/data directory. First the structures of the pairs are calculated using combfold. The best pairs are used to calculate higher order structures. Following this, the combfold algorithm is used to calculate the structure of the complex of the pairs and higher order pairs.

The colabfold step is done in paralel for each (higher order) pair. The outputs of colabfold are captured by re-evaluating the job-graph after the the pairings have been determined by Combfold.