Enrique Llobet Serra holds a degree in Biotechnology from the Catholic University of Valencia and a Master’s in Bioinformatics from the University of Valencia. His interest in artificial intelligence emerged during his master’s studies, during which he also completed an internship at MolDrug, developing QSAR models to predict the ADMET properties of small chemical molecules. Following this experience, he joined ProtoQSAR as a PhD candidate, working on the project “Modeling Materials Toxicity Using Artificial Intelligence-Based Techniques.” This initiative focuses on creating AI-based computational models to assess the safety of chemical substances and advanced materials, including nanomaterials.
OpenTox Summer School 2025
QSAR Models from Simple Chemicals to Advanced Materials
Llobet-Serra, Enrique1, Palomino-Schätzlein, Martina1,2, Vallés-Pardo, José Luis1; García-Meseguer, Rafael1, Serrano-Candelas, Eva1; Gozalbes, Rafael1,2
1ProtoQSAR SL, Nicolas Copérnico, 6, Parque Tecnológico de Valencia, Paterna 46980
2MolDrug AI Systems S.L., València (Spain)
Quantitative Structure-Activity Relationship (QSAR) models are an essential computational approach in fields like toxicology or drug discovery, enabling the prediction of a compound’s biological activity or toxicity from its molecular structure while reducing economic, ecological, and ethical burdens associated with traditional experiments. By minimizing cost and time, lowering reliance on animal testing, and allowing systematic investigation of extensive compound libraries, these models have gained widespread regulatory acceptance and proven applicability—particularly for discrete organic molecules, whose identification and characterization are highly advanced. Nanomaterials possess unique properties that enable extensive industrial applications. However, while they raise significant health and environmental concerns, far fewer tools are available to study their impact. Contrasting with the maturity of QSAR models for discrete organic molecules, nano-specific QSAR (nanoQSAR) approaches encounter various challenges, including limited data availability, complex material properties, and a lack of standardized protocols. Currently, existing nanoQSAR models primarily target specific substance classes (e.g., metal oxides, pure metals, or fullerenes), which enhance statistical reliability but limit broader predictive applicability. Therefore, the development of QSAR models tailored for nanomaterials and other advanced materials remains an active research area requiring significant effort. A particularly promising strategy is the exploration of novel descriptors beyond conventional molecular parameters, integrating experimental and electronic features to achieve more universal and accurate predictive models. Such advancements are critical for reliably predicting (eco)toxicological properties, improving material classification, and supporting effective grouping strategies for nanomaterials. The objective of this talk is to explain what QSAR models are, outline the workflow involved in developing QSAR models for simple molecules and discuss how these models can be adapted and applied to nanomaterials. In the practical session, participants will learn how to predict toxicological properties in a simple and efficient manner using QSAR models. ProtoPRED and other prediction tools will be presented, along with a series of practical exercises—including prediction of the mutagenicity of impurities (ICH-M7 Guidelines), predictions for REACH legislation, and evaluation of the ecotoxicological properties of compounds. Furthermore, the module ProtoNANO, specifically designed for (eco)toxicological predictions of nanomaterials, will be introduced and its application illustrated with several examples. This hands-on session is intended to provide essential knowledge and practical experience for the effective application of in silico models.