Acitores Cortina, J. M. | bioRxiv Science

bioRxiv2026

BioMADE: Predicting Torsades de Pointes from molecular structures through biologically informed representations

Drug-induced arrhythmias, particularly Torsades de Pointes (TdP), pose a significant risk to patient safety and can sometimes have life-threatening outcomes. They remain a major concern in drug development and regulation. Machine learning (ML) has become a powerful tool for analyzing complex biological and chemical datasets, enabling researchers to identify subtle patterns that differentiate safe compounds from those likely to cause dangerous cardiac effects. However, most existing in silico approaches do not sufficiently incorporate biological elements, relying heavily on chemical and structural properties or on computationally expensive simulations. Here, we introduce BioMADE, a novel ML framework that harnesses small-molecule-protein activity profiles from publicly available datasets to predict TdP risk without requiring exhaustive mechanistic annotation. Activity data from ChEMBL were used to train individual models for each gene, which predict activity values for any given compound. A curated set of arrhythmia-relevant genes was then used to construct a latent biological embedding (BioMADE embedding) for each molecule. We validated the performance of these features in distinguishing biological elements such as ATC3 class, showing superior classification performance compared with representations such as Molformer (lacks biological information) and MACCS (limited chemical properties) (0.85 AUROC vs 0.81 and 0.73, respectively). BioMADE representations served as input to a support vector machine classifier to discriminate TdP-inducing drugs from safe compounds. BioMADE achieved an AUROC of 0.89 in internal validation, indicating strong predictive performance. Against state-of-the-art models such as ADMEThyst, BioMADE achieved an AUROC of 0.74 on ADMEThysts validation set (vs. 0.72 for ADMEThyst). When we combined both approaches, the AUROC reached 0.77. These results demonstrate that BioMADE provides a scalable, biology-informed, and generalizable approach for predicting drug-induced toxicities. By integrating protein activity profiles into toxicology modeling, our framework highlights the critical role of human biology in adverse drug reaction prediction, an aspect often overshadowed by purely chemical or structural descriptors.

bioinformatics↗

bioRxiv2025

Cell-Type Specific Single-Cell Signatures Reveal Nephrotoxic Drug Effects

Drug-induced acute kidney injury (AKI) affects about 20% of hospitalized AKI patients, a significant contributor to morbidity and mortality. The lack of understanding of the kidney system and functioning of nephrotoxic drugs contributes to hospital-acquired AKI cases. AKI is difficult to predict because of its complex injury mechanism and the numerous pathways through which it manifests. Traditional toxicity biomarkers, like elevated creatinine levels, detect AKI only after significant kidney injury has occurred. Concurrently, advancements in single cell RNA sequencing (scRNAseq) have improved our ability to map cellular heterogeneity within tissues, potentially enabling the study of drug effects at a single cell level. We hypothesized that only particular subtypes of kidney cells may be responsible for observed nephrotoxicity and explain prediction challenges. To test this, we generated cellular response scores for 32 kidney cell types from the Human Cell Atlas and estimated drug effects. We identified significant expression differences in 6 cell types (e.g. Indistinct intercalated cell p = 0.009, Epithelial Progenitor cell, p = 0.04). We also developed an ensemble model that achieved an AUROC of 0.6 across different kidney cell populations - a significant improvement over using traditional bulk RNA sequencing alone. The single-cell transcriptomic signatures we identified potentially reveal unexplained molecular mechanisms of nephrotoxicity. Author SummaryThe prediction and early detection of drug-induced kidney injury is a significant clinical challenge since physicians rely on biomarkers that only become elevated after substantial kidney damage has occurred, limiting opportunities for intervention and patient protection. Our investigation utilized single-cell data and available drug toxicity information to examine how individual kidney cell populations respond to potentially harmful medications. We hypothesized that specific kidney cell subtypes are primarily responsible for observed drug toxicity, which may explain the difficulties in predicting drug-induced kidney injury. Through comprehensive analysis of 32 distinct kidney cell types, we identified six specific cellular populations that demonstrate differential responses to nephrotoxic compounds. We subsequently developed models that demonstrate superior predictive performance compared to analytical approaches using bulk RNA sequencing data. Our methodology represents a substantial advancement in precision medicine approaches to drug safety. These findings have important implications for clinical practice and patient safety. The cellular signatures we identified may enable earlier detection of kidney injury risk, potentially allowing clinicians to modify treatment regimens before irreversible damage occurs. Our work establishes a foundation for improved drug safety protocols and may contribute to reducing medication-related kidney injury in hospitalized patients.

bioinformatics↗