Software for processing large scale chemistry data
Welcome. This code is the result of a software engineering class project at the University of North Carolina at Chapel Hill. It was designed to compare sets of molecules to each other with high throughput by utilizing general purpose GPU computing.
This code will not run without a CUDA-capable GPU.
This package relies on the following and libraries:
After installing the dependencies, download this entire repository. All pyCUDA functions are contained in gpu_tanimoto_popcount.py. Copy this file into the same directory as the program that needs to use it.
To test that all the modules are working, run demo.py in the demo directory. This script compares 50,000 x 50,000 molecules.
Input molecules are formatted into their bitstring representations. For an example, see demo/bitstring-hex-encode.py
In sum, we made the speed improve by 5 in log scale in our benchmark.
