Pan-Assay Interference Compounds (PAINS) are nuisance compounds that interfere with biological assays leading to a false positive readout in drug discovery campaigns. PAINS filters are a set of expert-defined substructures commonly used in compound triage. While PAINS filters have garnered favor for their ease-of-use, transparency, and interpretability, these filters have shown to be overly sensitive. In order to provide a more data-driven assessment of PAINS, we have developed a web application that enables fast and informed profiling of compounds based on historical screening data. Our data includes ninety-seven thousand compounds, of which 16% have been flagged with PAINS filters. We provide the user with the information on the similarity of their compounds to compounds in our dataset and information for those compounds of assay the readings for the following assays: luciferase, beta-lactamase, and fluorescence. Our web application aims to present the pitfalls and dangers of relying solely on PAINS filters when triaging compounds. We present to users the assay results of compounds both with and without PAINS substructures. We have utilized the flask web tool package for the Python programming language to develop a web application. We also employ the RDKit library for substructure search, image generation, and the generation of fingerprints and Tanimoto similarity calculations. We have created an in-memory SQLite database with precomputed MACCS fingerprints, information on PAINS filters, and pre-rendered images to provide quick responses to users. The web app is freely available at https://pains.mml.unc.edu/. All the code is available at https://github.com/DnlRKorn/pains.

PAINS killer: A web application to evaluate the data driven validity of PAINS alerts

ACS Spring 2022

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