What question did this study set out to answer?

This research aims to develop an open-source QSAR pipeline for predicting pendrin inhibitors.

April 14, 20260 citationsOpen Access

pendrin-qsar: An Open-Source QSAR Pipeline for Pendrin (SLC26A4) Inhibitors

JKJoy KarmakarUniversity of California, San Francisco

Key Points

This research aims to develop an open-source QSAR pipeline for predicting pendrin inhibitors.
Developed a QSAR pipeline using machine learning techniques.
Curated dataset of 16 unique pendrin inhibitors from 6 scaffold classes.
Created four QSAR models: Random Forest, XGBoost, Ridge, and PLS.
Conducted virtual screening to identify new inhibitors, focusing on 13 benzofuran analogs.
Identified top virtual screening hit: V-2F4Cl with a predicted IC50 of 0.86 µM.
Validation techniques included LOO-CV, Y-scrambling, and applicability domain analysis.
QSAR models demonstrated robust predictive performance with comprehensive validation.

Abstract

An open-source cheminformatics and machine learning pipeline for pendrin (SLC26A4) inhibitors. Includes a curated dataset of 16 unique inhibitors across 6 scaffold classes, four QSAR models (Random Forest, XGBoost, Ridge, PLS) validated by LOO-CV, Y-scrambling, applicability domain analysis, and conformal prediction, plus virtual screening of 13 novel benzofuran analogs. Top predicted virtual screening hit: V-2F4Cl (predicted IC50 = 0.86 µM, 80% CI: 0.43-1.73 µM). Code runs in a single click in Google Colab with no installation required. GitHub: https://github.com/drjoykarmakar/pendrin-qsar

问 AI

Bookmark

View Full Paper