Artificial Intelligence in Drug Toxicity Prediction: Recent Advances, Challenges, and Future Perspectives

Abstract

Toxicity prediction is a critical step in the drug discovery process that helps identify and prioritize compounds with the greatest potential for safe and effective use in humans, while also reducing the risk of costly late-stage failures. It is estimated that over 30% of drug candidates are discarded owing to toxicity. Recently, artificial intelligence (AI) has been used to improve drug toxicity prediction as it provides more accurate and efficient methods for identifying the potentially toxic effects of new compounds before they are tested in human clinical trials, thus saving time and money. In this review, we present an overview of recent advances in AI-based drug toxicity prediction, including the use of various machine learning algorithms and deep learning architectures, of six major toxicity properties and Tox21 assay end points. Additionally, we provide a list of public data sources and useful toxicity prediction tools for the research community and highlight the challenges that must be addressed to enhance model performance. Finally, we discuss future perspectives for AI-based drug toxicity prediction. This review can aid researchers in understanding toxicity prediction and pave the way for new methods of drug discovery.

A Promising Therapeutic Scaffold Fusing Chalcone/Coumarin Targeting Colorectal Cancer: Design, Synthesis, Biological and ADME‐tox Modelling

“…In addition to pharmacokinetic studies, twelve toxicological endpoints have also been estimated for the lead‐hybrid 3 c by employing several open‐sources chemoinformatic servers such as OSIRIS, TEST, ProTox‐II, Pred‐hERG, pkCSM, SwissADME, ToxTree, and ADMET‐SAR (Table 3). These parameters are closely associated to adverse effects in the progress of a lead molecule, and their early estimation through artificial intelligence tools offer an attractive and rapid low‐cost approach toward the design of safer pharmaceutical lead candidates [75,76] . Regarding their in silico toxicological diagnosis, promising 3 c would have no apparent warnings, precautions and adverse events as mutagenic, tumorigenic, irritant, hepato/nephro/neuro/cardio/or inmunotoxic.…”

Benchmarking of Small Molecule Feature Representations for hERG, Nav1.5, and Cav1.2 Cardiotoxicity Prediction

“…Compounds were classified based on their IC50 values, following standard criteria used by researchers in the field. ,, Compounds with IC50 values of 10 μM or below (pIC50 ≥ 5) were categorized as blockers (inhibitors), while compounds with IC50 values higher than 10 μM (pIC50 < 5) were categorized as nonblockers (inactive).…”

“…This transition is primarily due to their superior empirical performance in the field, as evidenced by multiple research publications outlined recently. 25 However, there is still important room for improvement in cardiotoxicity prediction. First, almost all published methods during the last 20 years focused only on hERG liability prediction, as there were very few bioactivity data available on the other two cardiac ion channels.…”

“…Later, there has been a notable shift toward the utilization of random forest (RF), support vector machine (SVM), and deep neural network (DNN) methods, − using up to 15,000 compounds for training the models. This transition is primarily due to their superior empirical performance in the field, as evidenced by multiple research publications outlined recently …”

Recent Studies of Artificial Intelligence on In Silico Drug Absorption

“…However, we only focus on highlighting 11 properties that have been of interest to many AI-based ADMET researchers, including the logarithm of the octanol–water partition coefficient (log P), the logarithm of pH-dependent distribution coefficient (log D), the logarithm of the aqueous solubility (log S), p K a , human oral bioavailability (HOB), human intestinal absorption (HIA), Caco-2 cell permeability, P-glycoprotein (P-gp) inhibitor and substrate, parallel artificial membrane permeability assay (PAMPA), and Madin-Darby Canine Kidney Cells (MDCK) permeability. Interested readers can refer to other useful ADMET property reviews. − …”