hERG-LTN: A New Paradigm for hERG Cardiotoxicity Prediction Using Neuro-Symbolic and Generative AI

Assessing adverse drug reactions (ADRs) during drug development is essential for ensuring the safety of new compounds. The blockade of the Ether-à-go-go-related gene (hERG) channel plays a critical role in cardiac repolarization. Computational predictions of hERG inhibition can help foresee drug safety, but current data-driven approaches have limitations. Therefore, a new paradigm that bridges the gap between data and knowledge offers an alternative for advancing precision pharmacogenomics in assessing hERG cardiotoxicity. This study aims to develop a reasoning-based, in-silico, robust model for predicting drug-induced hERG inhibition, facilitating new drug development by reducing time and cost, and supporting downstream in-vitro and in-vivo testing. In this study, we constructed a new dataset (UnihERG_DB) by sourcing data from ChEMBL, PubChem, BindingDB, hERG Karim's, hERG Blocker's, and GTP bioactivity databases. The final dataset comprises 20,409 SMILES (Simplified Molecular Input Line Entry System) data samples, labeled as hERG blockers (IC50 < 10 μM) or non-hERG blockers (IC50 ≥ 10 μM). Molecular features were extracted using Morgan and CDK fingerprints. Furthermore, we explored embedding feature computation using cutting-edge Large Language Models, including NVIDIA MegaMolBART, LLaMA 3.2, Gemini, and DeepSeek. Finally, we utilized the Logic Tensor Network (LTN), an advanced AI framework, to train and develop the hERG predictive model. Model performance was evaluated using two benchmarks: External Test-1 and hERG-70. The model outperformed several SOTA approaches, including CardioTox, M-PNN, DeepHIT, CardPred, OCHEM Predictor-II, Pred-hERG 4.2, Random Forest, and Gradient Boosting. On the External Test-1 dataset, hERG-LTN achieved an accuracy (ACC) of 0.931, a specificity (SPE) of 0.928, and a sensitivity (SEN) of 0.933. Furthermore, on the hERG-70 benchmark, our method achieved an accuracy of 0.827, a specificity of 0.890, and a correct classification rate (CCR) of 0.833. Overall, the Neuro-Symbolic AI approach sets a new standard for hERG-related cardiotoxicity assessment, yielding competitive results with current state-of-the-art (SOTA) models, and highlights its potential for advancing precision pharmacogenomics in drug discovery and development (GitHub: https://github.com/hossain013/hERG-LTN).