Spatial artefact detection improves reproducibility of drug screening experiments

Abstract

Reliable and reproducible drug screening experiments are essential for drug discovery and personalized medicine. Here, we demonstrate how systematic experimental errors negatively impact reproducibility, and that conventional quality control (QC) methods based on plate controls fail to detect these errors. To address this limitation, we developed a control-independent QC approach using normalized residual fit error (NRFE) to identify systematic errors in drug screening experiments. Comprehensive analysis of >100,000 duplicate measurements from the PRISM pharmacogenomic study revealed that the NRFE-flagged experiments show three-fold lower reproducibility between technical replicates. By integrating NRFE with existing QC methods to analyze 41,762 matched drug-cell line pairs between two datasets from the Genomics of Drug Sensitivity in Cancer project, we improved the cross-dataset correlation from 0.66 to 0.76. Available as an R package at https://github.com/IanevskiAleksandr/plateQC, plateQC provides a robust toolset for enhancing drug screening data reliability and consistency for basic research and translational applications.