NanoToxRadar: A Multitarget Nano-QSAR Model for Predicting the Cytotoxicity Values of Multicomponent Nanoparticles

Area of Science:

• Nanotechnology
• Toxicology
• Computational Chemistry

Background:

• Nanotechnological advancements have yielded complex nanoparticles, necessitating robust toxicity assessment tools.
• Existing nano-QSAR models have limited applicability domains, restricting their use to specific nanoparticle types and cell lines.
• There is a need for predictive models that can assess the toxicity of diverse multicomponent nanoparticles (MC-NPs) across various cell types.

Purpose of the Study:

• To develop NanoToxRadar, a web-based platform for predicting the toxicity of MC-NPs toward a wide range of cell lines.
• To extend the applicability domain (AD) of nano-quantitative structure-activity relationship (nano-QSAR) models for complex nanomaterials.
• To provide an accessible tool for researchers to estimate nanoparticle toxicity.

Main Methods:

• Development of a web-based platform (NanoToxRadar) integrating a predictive toxicity model.
• Representation of MC-NP molecular structures using a size-dependent electron-configuration fingerprint.
• Utilization of one-hot encoded cell types and a CatBoost regression model to predict toxicities against 110 cell lines.

Main Results:

• The CatBoost regression model demonstrated strong predictive performance with an R²_Test of 0.877.
• NanoToxRadar successfully predicts pIC₅₀ values for 110 cell lines based on nanoparticle composition (core, shell, dopant, coating) and diameter.
• The platform is accessible online at https://www.kitox.re.kr/nanotoxradar.

Conclusions:

• NanoToxRadar effectively predicts the toxicity of multicomponent nanoparticles across a broad spectrum of cell lines.
• The developed platform significantly expands the applicability domain of nano-QSAR models for complex nanomaterials.
• NanoToxRadar serves as a valuable resource for assessing nanoparticle safety and guiding future research in nanomedicine and nanotoxicology.