Saarimäki Laura Aliisa, Fratello Michele, Del Giudice Giusy, Di Lieto Emanuele, Afantitis Antreas, Alenius Harri, Chiavazzo Eliodoro, Gulumian Mary, Karisola Piia, Lynch Iseult, Mancardi Giulia, Melagraki Georgia, Netti Paolo, Papadiamantis Anastasios G, Peijnenburg Willie, A Santos Hélder, Serchi Tommaso, Shahbazi Mohammad-Ali, Stoeger Tobias, Valsami-Jones Eugenia, Vivo Paola, Vinković Vrček Ivana, Vogel Ulla, Wick Peter, Winkler David A, Serra Angela, Greco Dario
Finnish Hub for Development and Validation of Integrated Approaches (FHAIVE), Faculty of Medicine and Health Technology, Tampere University, Tampere 33520, Finland.
Division of Pharmaceutical Biosciences, Faculty of Pharmacy, University of Helsinki, Helsinki 00790, Finland.
Environ Sci Technol. 2025 Jul 29;59(29):14969-14980. doi: 10.1021/acs.est.5c00841. Epub 2025 Jul 17.
The development of new approach methodologies (NAMs) to replace current testing for the safety assessment of engineered nanomaterials (ENMs) is hindered by the scarcity of validated experimental data for many ENMs. We introduce a framework to address this challenge by harnessing the collective expertise of professionals from multiple complementary and related fields ("wisdom of crowds" or WoC). By integrating expert insights, we aim to fill data gaps and generate consensus concern scores for diverse ENMs, thereby enhancing the predictive power of nanosafety computational models. Our investigation reveals an alignment between expert opinion and experimental data, providing robust estimations of concern levels. Building upon these findings, we employ predictive machine learning models trained on the newly defined concern scores, ENM descriptors, and gene expression profiles, to quantify potential harm across various toxicity end points. These models further reveal key genes potentially involved in underlying toxicity mechanisms. Notably, genes associated with metal ion homeostasis, inflammation, and oxidative stress emerge as predictors of ENM toxicity across diverse end points. This study showcases the value of integrating expert knowledge and computational modeling to support more efficient, mechanism-informed, and scalable safety assessment of nanomaterials in the rapidly evolving landscape of nanotechnology.
用于替代当前工程纳米材料（ENM）安全性评估测试的新方法学（NAM）的发展，受到许多ENM缺乏经过验证的实验数据的阻碍。我们引入了一个框架，通过利用来自多个互补和相关领域的专业人员的集体专业知识（“群体智慧”或WoC）来应对这一挑战。通过整合专家见解，我们旨在填补数据空白，并为各种ENM生成共识关注分数，从而提高纳米安全计算模型的预测能力。我们的调查揭示了专家意见与实验数据之间的一致性，为关注水平提供了可靠的估计。基于这些发现，我们使用在新定义的关注分数、ENM描述符和基因表达谱上训练的预测性机器学习模型，来量化各种毒性终点的潜在危害。这些模型进一步揭示了可能参与潜在毒性机制的关键基因。值得注意的是，与金属离子稳态、炎症和氧化应激相关的基因成为不同终点ENM毒性的预测指标。这项研究展示了整合专家知识和计算建模的价值，以支持在快速发展的纳米技术领域中对纳米材料进行更高效、基于机制且可扩展的安全性评估。
