Benchmarking graph-based models for in-silico toxicity prediction in drug discovery
Abstract
Drug discovery is a costly and high-risk process, where toxicity-related failures remain a major cause of attrition in both preclinical and clinical stages. As a result, accurate early prediction of chemical toxicity is essential to reduce downstream costs and improve compound prioritization. In this context, graph deep learning (GDL) has emerged as a powerful paradigm for toxicity prediction, leveraging molecular graph representations to learn directly from chemical structure with improved expressivity over traditional approaches. Despite the growing number of proposed models, current literature-based comparisons are often difficult to interpret due to inconsistencies in datasets, preprocessing pipelines, and evaluation protocols. To address this limitation, we introduce a unified and standardized benchmarking framework for GDL-based toxicity prediction. We systematically evaluate more than 20 representative approaches under consistent experimental conditions and across multiple datasets and partitioning strategies, enabling a fair and reproducible comparison of model performance. In addition, we complement this empirical study with a structured literature analysis to contextualize existing methodological trends and performance claims. Our results provide a clearer and more reliable assessment of the current state of the field, highlighting both the strengths and limitations of existing graph-based approaches. To support transparency and reproducibility, we release our benchmarking framework as open-source software https://gitlab.citius.gal/noel.suarez/benchtox, allowing the community to evaluate and compare models under consistent conditions.
Related
- Benchmarking Test Statistics and False Discovery Rate Procedures for Single-Cell Differential Expression
- Benchmarking AI scientists for omics data–driven biological discovery
- Machine Learning for Toxicity Prediction in Low-Sample Molecular Classes
- Gut Microbiome Alterations in TAAR5 Knockout Mice and In Silico Ligand Prediction.
- From in silico prediction to experimental validation: Identification of drugs and novel synergistic combinations that inhibit growth of inflammatory breast cancer cells
BibTeX
@misc{suarezbarro2026benchmarking,
title = {{Benchmarking graph-based models for in-silico toxicity prediction in drug discovery}},
author = {Noel Suarez-Barro and Manuel Lama and Juan C. Vidal},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.37555},
eprint = {2609.37555},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.37555},
}
RIS
TY - UNPB
TI - Benchmarking graph-based models for in-silico toxicity prediction in drug discovery
AU - Noel Suarez-Barro
AU - Manuel Lama
AU - Juan C. Vidal
PY - 2026
DA - 2026/09/29
PB - arXiv
DO - 10.48550/arxiv.2609.37555
UR - https://arxiv.org/abs/2609.37555
AB - Drug discovery is a costly and high-risk process, where toxicity-related failures remain a major cause of attrition in both preclinical and clinical stages. As a result, accurate early prediction of chemical toxicity is essential to reduce downstream costs and improve compound prioritization. In this context, graph deep learning (GDL) has emerged as a powerful paradigm for toxicity prediction, leveraging molecular graph representations to learn directly from chemical structure with improved expressivity over traditional approaches. Despite the growing number of proposed models, current literature-based comparisons are often difficult to interpret due to inconsistencies in datasets, preprocessing pipelines, and evaluation protocols. To address this limitation, we introduce a unified and standardized benchmarking framework for GDL-based toxicity prediction. We systematically evaluate more than 20 representative approaches under consistent experimental conditions and across multiple datasets and partitioning strategies, enabling a fair and reproducible comparison of model performance. In addition, we complement this empirical study with a structured literature analysis to contextualize existing methodological trends and performance claims. Our results provide a clearer and more reliable assessment of the current state of the field, highlighting both the strengths and limitations of existing graph-based approaches. To support transparency and reproducibility, we release our benchmarking framework as open-source software https://gitlab.citius.gal/noel.suarez/benchtox, allowing the community to evaluate and compare models under consistent conditions.
ER -