In recent years, the cheminformatics community has seen an increased success with machine learning-based scoring functions for estimating binding affinities and pose predictions. By their very nature, the training and evaluation of these types of scoring functions requires the processing of large amounts of data, which includes the integration of several tools for handling the conversion between several chemical file formats, compilation of results, and the application of functions for rigorous statistical analysis. In order to enhance the reproducibility and reliability of our research, we found that it was necessary to develop a workflow that combined all of these steps. Here we present a prototype pipeline, and our results from analyzing a novel deep learning scoring function, DLScore, which already shows promising results when compared to existing scoring functions, such as nnscore. In the future, we plan to integrate this workflow with existing platforms and tools, such as the Drug Design Data Resource (D3R) workflows, the Open Drug Discovery Toolkit (oddt), and the DrugDiscovery@TACC portal.

Automated workflow for reproducible analysis of protein-ligand scoring functions

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.