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

Abstract

In the field of drug discovery, there is a substantial challenge in seeking out chemical structures that possess desirable pharmacological, toxicological, and pharmacokinetic properties. Complications arise when drugs interfere with the functioning of cardiac ion channels, leading to serious cardiovascular consequences. The discontinuation and removal of numerous approved drugs from the market or at late development stages in the pipeline due to such inhibitory effects further highlight the urgency of addressing this issue. Consequently, the early prediction of potential blockers targeting cardiac ion channels during the drug discovery process is of paramount importance. This study introduces a deep learning framework that computationally determines the cardiotoxicity associated with the voltage-gated potassium channel (hERG), the voltage-gated calcium channel (Cav1.2), and the voltage-gated sodium channel (Nav1.5) for drug candidates. The predictive capabilities of three feature representationsmolecular fingerprints, descriptors, and graph-based numerical representationsare rigorously benchmarked. Additionally, a novel training and evaluation data set framework is presented, enabling predictive model training of drug off-target cardiotoxicity using a comprehensive and large curated data set covering these three cardiac ion channels. To facilitate these predictions, a robust and comprehensive small molecule cardiotoxicity prediction tool named CToxPred has been developed. It is made available as open source under the permissive MIT license at .

Semi-Supervised Learning to Boost Cardiotoxicity Prediction by Mining a Large Unlabeled Small Molecule Dataset

“…In the labeled set, compounds exhibiting inhibitory activity were collected from multiple public data repositories, including the ChEMBL bioactivity database [28][29][30], PubChem [31], BindingDB [32][33], hERGCentral [34], and US patent and literature-derived data [35][36][37][38]. Further information on the manual data curation process can be accessed in [20].…”

“…The calculation of molecular descriptors was performed using the Mordred Python package [50]. We employed 2D descriptors only as these require fewer computational resources compared to 3D descriptors without sacrificing predictive performance [20][51][52][53], resulting in a total of 1613 descriptors. Preprocessing and feature selection were accomplished through a Scikit-Learn [54] pipeline consisting of four modules.…”

“…Several previous reviews [14][17][18][19] have underscored this shift, also highlighting a notable observation in recent ML methods presented, favoring the use of random forest (RF), support vector machine, and deep neural network methods, primarily due to their demonstrated superior empirical performance. As stated in our previous research [20], there exists significant room for improvement in the field of cardiotoxicity prediction. Such improvement includes expanding computational methods beyond solely focusing on hERG, but to also tackle the other crucial cardiac ion channels: Nav1.5 and Cav1.2.…”

“…This study builds on our previous work [20] by mining the whole database of small molecules in ChEMBL, comprising approximately 2 million compounds. With this approach, our goal was to construct cardiotoxicity classifiers that are robust, reliable, and possess superior generalization capabilities.…”

Editorial: Machine Learning in Bio-cheminformatics

“…Some methods, algorithms, and functional tools were constructed to facilitate the application or improve the performance of the classic virtual screening strategy. − Machine learning methods were also adopted in this collection to identify new hit compounds, discover promising leads for cholestasis, interpret QSAR models, and learn molecular representations. − DiStefano et al and Mao et al conducted research on toxicity prediction and antiviral drug design, respectively. Moreover, ML was also applied to explore the pharmaceutical properties of diverse drug candidates. − A novel knowledge base for nonalcoholic fatty liver disease was developed . Heyndrickx et al adopted cross-pharma federated learning to unleash the benefit of QSAR.…”