NanoRisk-Web: An open-source Bayesian decision-support platform for occupational health risk assessment of engineered nanomaterials

Abstract

Engineered nanomaterials (ENMs) may pose occupational health risks during handling and application. NanoRisk-Web is an open-source web-based decision-support platform that integrates Bayesian Network modelling with an enhanced control-banding framework for ENM risk assessment. The software predicts hazard levels from physicochemical descriptors, evaluates exposure potential using process-specific parameters, and quantifies overall risk to recommend appropriate control measures. Developed in Python and hosted on Streamlit, NanoRisk-Web incorporates an exact Bayesian inference engine that reproduces the posteriors of the licensed SMILE and GeNIe engines (BayesFusion) while removing proprietary runtime dependencies. Software verification showed maximum absolute differences below 1e -6 across the tested node marginals. The embedded Bayesian Network was additionally evaluated on an independent held-out set of 91 records, achieving 91.21% overall classification accuracy. By enabling scenario-based probabilistic risk assessment of nanomaterials, the platform is intended as a transparent and reproducible decision-support implementation rather than a substitute for workplace exposure measurement or professional occupational-hygiene judgement