Integrating multi-type features and knowledge graph for graded prediction of drug-induced liver injury in humans
Ying Liu, Kaimiao Hu, Jie Geng, Qi Dai, Leyi Wei, Ran Su
Abstract
Drug-induced liver toxicity poses a threat to human health and remains a significant reason for drug withdrawal from the market. Therefore, early identification of drug-induced liver injury (DILI) during drug development is crucial. However, most studies on hepatotoxicity prediction are limited to single type of features or binary toxicity assessment. In this study, we propose a novel liver toxicity prediction model called MolFPKG-DILI (Molecular Graph, FingerPrint and Knowledge Graph-based DILI), which integrates multi-type compound features and knowledge graph for assessing DILI severity.
Introduction
Drug-induced liver injury (DILI) is a prevalent adverse reaction in clinical practice, ranging from mild elevation of liver enzymes to severe liver dysfunction and even liver failure, posing a significant threat to life. Studies have shown that between 1990 and 2010, a total of 133 drugs were removed from the market due to safety concerns, approximately 27.1% of which were due to liver toxicity. Therefore, it is crucial to assess the potential hepatotoxicity risk of drugs during the clinical development and application of drugs. Researchers commonly employ in vivo and in vitro experiments as well as animal studies to assess liver toxicity.
Materials and Methods:
We collected drugs related to DILI from the DILIrank and DILIst databases. In DILIrank, compounds are categorized based on their DILI risk into four classes: Most-DILI-concern (192 drugs), Less-DILI-concern (278 drugs), No-DILI-concern (312 drugs), and Ambiguous DILI-concern (254 drugs). DILIst classifies compounds into two categories: DILI positives (768 drugs) and DILI negatives (511 drugs). We initially selected DILI-negative drugs from the DILIst, along with Most-DILI-concern and Less-DILI-concern drugs from the DILIrank. Subsequently, drugs without retrievable SMILES were excluded, resulting in 477 No-DILI and 453 DILI (including 182 Most-DILI and 271 Less-DILI).
Discussion:
To gain a comprehensive understanding of the compounds involved in this study, we conducted data analysis and visualization of both toxic and non-toxic drugs. The Tanimoto coefficient is commonly used to measure the similarity between compounds, and we utilized this coefficient to generate a similarity heatmap of the compounds. It primarily appears yellow, indicating low similarity among the compounds in the dataset and highlighting the chemical structural diversity. We employed the t-SNE (t-Distributed Stochastic Neighbor Embedding) algorithm to map the high-dimensional drug Morgan fingerprints into a three-dimensional space for drug visualization.
Citation: Liu Y, Hu K, Geng J, Dai Q, Wei L, Su R (2026) Integrating multi-type features and knowledge graph for graded prediction of drug-induced liver injury in humans. PLoS Comput Biol 22(7): e1013640. https://doi.org/10.1371/journal.pcbi.1013640
Editor: Roger Dimitri Kouyos, University of Zurich, Division of Infectious Diseases and Hospital Epidemiology, Raemistrasse 100, SWITZERLAND, Zurich, ZH, 8091
Received: January 24, 2025; Accepted: October 21, 2025; Published: July 14, 2026.
Copyright: © 2026 Liu 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 used for running experiments and model fitting is available on a GitHub repository at https://github.com/RanSuLab/MolFPKG-DILI.
Funding: This study was supported by the National Natural Science Foundation of China (Grant No. 62222311 to RS) and Tianjin Science and Technology Plan Project (Grant No. 22JCZDJC00580 to JG). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
