A genomic-led strategy to anticipate drug safety effects
Brian R. Ferolito, Andrea R. V. R. Horimoto, Kai Gravel-Pucillo, Daniel J. Golden, Hesam Dashti, Claudia Giambartolomei, Danielle Rasooly, Rachael Matty, Liam Gaziano, Yakov Tsepilov, Lauren Costa, Nicole Kosik, Harris Ioannidis, Mohd Karim, Giovanna Winicki, Fiona Hunter, Claudia Langenberg, John C. Whittaker, Million Veteran Program, Tianxi Cai, Gina M. Peloso, Barbara Zdrazil, Maya Ghoussaini, Andrew R. Leach, Sumitra Muralidhar, Ines A. Smit, Juan P. Casas, J. Michael Gaziano, Kelly Cho, Alexandre C. Pereira
Abstract
Safety-related issues account for approximately 25% of failures in new drug discovery programs. On top of that, many are discovered during post-marketing surveillance, significantly limiting drug utility and application. To proactively address these concerns, we developed a genetics-led strategy leveraging Mendelian Randomization (MR) across large-scale genetic datasets from the Million Veteran Program, FinnGen, and UK Biobank. By mapping genetic variants associated with gene expression and protein abundance to 1,449 harmonized human phenotypes, we systematically identified potential adverse drug reactions (ADR).
Introduction
Safety-related issues are responsible for approximately 25% of new drug discovery program failures, and even among drugs that reach approval, unexpected adverse drug reactions (ADRs) often surface only after widespread clinical use [1,2]. The development of a framework to predict adverse drug reactions (ADR) could substantially increase the success rate of drug development programs as well as the identification and reporting of ADRs for novel and existing drugs [3].
Materials and method
Experimental design
GWAS summary statistics were obtained from the Million Veteran Program [42] (MVP), FinnGen [43] R.10, and the UK Biobank [44,45] (UKBB). From UKBB and MVP, we included 1,556 unique Phecodes, 72 biomarkers, 74 questionnaire data, and 6 clinical variables. A total of 1,449 traits were harmonized between datasets and meta-analyzed. For meta-analyzed traits, we performed fixed effects inverse-variance weighted meta-analysis using METAL [46].
Results
A catalog of potential adverse drug events
The two-sample MR filtering strategy and the analysis flowchart are represented in Fig 1. We started with 58,276 gene-trait pairs that passed the threshold for MR significance (p-value < 1.59x10-9). From these we mapped the predicted MoA that would lead to an adverse event (e.g., a positive or negative modulator). Briefly, we used the MR-beta directionality and the safety signal metrics (defined in Methods; Table A in S1 Table) to develop an atlas (Table A in S1 Table), in which, given a gene and a pharmacological action of a particular drug, we describe all of the genetically predicted safety concerns for that specific modulation of the target gene.
Discussion
This study presents a systematic approach to leveraging Mendelian randomization (MR) and genetic evidence for identifying adverse drug reactions (ADR) associated with gene-target modulations. By integrating data on gene-trait relationships with pharmacological mechanisms, we developed a comprehensive safety atlas that predicts potential adverse outcomes of therapeutic interventions. This atlas represents a significant advancement in drug safety research, contributing to both drug development and post-market safety assessments.
Acknowledgments
Full Million Veteran Program acknowledgement can be found in S3 Text.
We would like to acknowledge the time and effort of the study participants and researchers in the Fenland study https://doi.org/10.22025/2017.10.101.00001; https://www.mrc-epid.cam.ac.uk/research/studies/fenland/.
Citation: Ferolito BR, Horimoto ARVR, Gravel-Pucillo K, Golden DJ, Dashti H, Giambartolomei C, et al. (2026) A genomic-led strategy to anticipate drug safety effects. PLoS Genet 22(7): e1012211. https://doi.org/10.1371/journal.pgen.1012211
Editor: Stuart A. Scott, Stanford University, UNITED STATES OF AMERICA
Received: September 30, 2025; Accepted: June 8, 2026; Published: July 16, 2026
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: Full summary statistics of all the two-sample MR results used in this analysis are publicly available and can be downloaded at the Department of Veterans Affairs Centralized Interactive Phenomics Resource (CIPHER) web portal (https://phenomics.va.ornl.gov/) or at OSF (https://doi.org/10.17605/OSF.IO/G8HJZ) https://osf.io/g8hjz/overview [55].
Funding: This research is based on data from the Million Veteran Program, Office of Research and Development, Veterans Health Administration, and was supported by award #MVP000 (KC, MG, GW, DG, BF, AP, KGP, RM, LC, NK). This publication does not represent the views of the Department of Veteran Affairs or the United States Government. This research used resources from the Knowledge Discovery Infrastructure at the Oak Ridge National Laboratory, supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725 (KC, MG, GW, DG, BF, AP, KGP, RM, LC, NK) and the Department of Veterans Affairs Office of Information Technology Inter-Agency Agreement with the Department of Energy under IAA No. VA118-16-M-1062. JCW is funded by the UK Medical Research Council via programme grant MC_UU_00002/18. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Maya Ghoussaini is a full-time employee at Regeneron Genetics Centre. Maya’s main contributions were while she was an employee of Open Targets. Mohd Karim is a full-time employee at Variant Bio. Mohd’s main contributions were while he was a an employee of Open Targets. JP Casas is a full-time employee at Novartis Institutes for Biomedical Research. JP Casas’ main contributions to the project were while employed at the VA Boston Healthcare System.
