HPC Batch Processing for KinasePredictPro (KIPP)

This fork adds high-performance computing (HPC) capabilities to KinasePredictPro for processing large libraries in parallel using SLURM clusters.

##Modifications

###Problem The original code accesses the same models directory causing I/O bottleneck when running hundreds of parallel jobs.

Solution

Batch processing mode: Process chunk of SMILES (Each SLURM array gets 1 chunk) Model pre-loading: Load models once per job, share across workers (Each SLURM job copies models to temp file, deletes after) Python muliprocessing: Switched from parallel to python internal multiprocessing with nesteed parallelism

Quickstart

Original project

For General KIPP usage and citation inforamtion see README.md

#Author Jeremy Leitz - HPC modifications 2025