Systematic Identification of Pan-cancer Single-gene Expression Biomarkers in Drug High-throughput Screens
Ginte Kutkaite, Göksu Avar, Diyuan Lu, Thomas J. O’Neill, Daniel Krappmann, Michael P. Menden
Abstract
Precision oncology relies on molecular biomarkers to stratify patients into responders and non-responders to a given treatment. Although gene expression profiles have historically been explored for biomarker discovery, fewer studies investigated single-gene expression biomarkers. Additionally, many approaches are limited to cancer type-specific associations, which constrain statistical power. To address these limitations, we developed a regression-based framework that corrects for tissue-specific biases and enhances detection of pan-cancer single-gene expression biomarkers of drug sensitivity in cancer cell line high-throughput drug screens.
Introduction
Precision oncology seeks to improve treatment outcomes by stratifying patients based on their molecular profiles to predict therapeutic response. Despite advances in molecular profiling technologies, drug development remains high-risk, with clinical trial failure rates nearing 95% often due to the absence of reliable biomarkers for identifying responsive subgroups. This underscores the urgent need for novel biomarkers and innovative application strategies to accelerate drug development and improve clinical success.
Materials and Methods:
To extend the feature analysis, we next examined how tissue correction influences the composition and tissue dependence of top-ranked genes across informative drug models. We evaluated the effect of tissue correction on feature selection using the top 10 ranked genes from each model trained on raw and tissue-z-score-corrected gene expression matrices. We classified genes as retained when present in both raw and corrected models, and as emerged when present only after correction.
Discussion
Genomic profiling within individual cancer types has driven early success in precision oncology by enabling targeted therapies against recurrent oncogenic mutations. However, progress has slowed due to tumor heterogeneity, limited cohort sizes, and the rarity of actionable mutations, all of which constrain predictive modeling and clinical translation. In contrast, gene expression (GEX) profiling and pan-cancer analyses remain underutilized, despite their potential to capture functional tumor states and offer increased statistical power. Harnessing these complementary data layers presents a key opportunity to accelerate progress in precision oncology.
Acknowledgments
We are grateful for the valuable discussions with colleagues at Helmholtz Munich and the support from our funding agencies.
Citation: Kutkaite G, Avar G, Lu D, O’Neill TJ, Krappmann D, Menden MP (2026) Systematic identification of pan-cancer single-gene expression biomarkers in drug high-throughput screens. PLoS One 21(5): e0330412. https://doi.org/10.1371/journal.pone.0330412
Editor: UDAYAN BHATTACHARYA, Weill Cornell University, UNITED STATES OF AMERICA
Received: July 31, 2025; Accepted: April 8, 2026; Published: May 11, 2026.
Copyright: © 2026 Kutkaite et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data and code are available as part of Supporting Information associated with the manuscript and via the GitHub repository (https://github.com/MendenLab/Pan-can_GEX_biomarkers).
Funding: The research by M.P.M. is supported by a H2020 European Research Council (ERC) grant (agreement No. 950293). D.K. is supported by Deutsche Krebshilfe (grant 70115440). Funders did not play any role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: M.P.M. collaborates with and receives funding from AstraZeneca, GSK and F. Hoffmann-La Roche. M.P.M. also consults for MSD and McKinsey. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
