Abstract

IN SILICO TOXICITY SCREENING: FROM CONVENTIONAL TO DIGITAL TOXICOLOGY

Saira Bhanu Shaik*, M. Vamsi Vara Prasad, M. A. V. D. S. Prajna, P. Sai Srinu, A. Vara Lakshmi, Ch. Siva Durga, G. Akhila Bhanu Prasanna, Ch. Janush

Abstract

Toxicity remains a major consideration in drug discovery because a compound must combine therapeutic activity with an acceptable safety profile. Conventional toxicity testing provides essential biological evidence but can be resource-intensive and may be difficult to apply to large chemical libraries at an early stage. In silico toxicity screening provides a complementary, data-driven approach for identifying potential safety liabilities before extensive experimental testing. This review summarises the transition from conventional toxicology to digital toxicology and discusses quantitative structure–activity relationship (QSAR) modelling, read-across, molecular docking, toxicokinetic modelling, machine learning, deep learning and artificial intelligence (AI) approaches. Major toxicity endpoints, including cardiotoxicity, hepatotoxicity, nephrotoxicity, neurotoxicity, genotoxicity, carcinogenicity, reproductive and developmental toxicity, immunotoxicity and respiratory toxicity, are considered from a computational perspective. The article also examines how computational methods can support ADMET profiling, lead optimisation, candidate prioritisation and early toxicity-based deprioritisation. Although digital methods can improve speed, scalability and consistency, their reliability depends on data quality, model validation, applicability domain, interpretability and experimental confirmation. Future toxicology is expected to increasingly integrate AI, multi-omics, human-relevant in vitro systems, organ-on-chip models and New Approach Methodologies (NAMs). Overall, computational toxicity assessment is best viewed as a complementary component of an integrated safety strategy rather than a replacement for experimental toxicology.

Keywords: in silico toxicology; toxicity screening; QSAR; read-across; machine learning; artificial intelligence; ADMET; drug discovery; computational toxicology.