Unravelling PFAS disruption of thyroxine transport by transthyretin using machine learning-augmented modelling

Authors: Theodori, Aikaterini; Papavasileiou, Konstantinos D; Tzoupis, Haralampos; Zareifi, Danae; Tsoumanis, Andreas; Dondero, Francesco; Melagraki, Georgia; Lynch, Iseult; Afantitis, Antreas

Source journal: Journal of hazardous materials (2026-11-12)

DOI: 10.1016/j.jhazmat.2026.144009

Abstract

Thyroid-disrupting chemicals (TDCs) pose a significant risk to human health and are associated with adverse neurological and cardiovascular effects. Per- and polyfluoroalkyl substances (PFAS) have been shown to bind to transthyretin (TTR), disrupting thyroid hormone homeostasis and leading to adverse outcomes. We present a refined workflow for modelling and predicting PFAS activity toward TTR, where training compounds were labelled Active or Inactive based on experimental activity values (binding affinities) exceeding 50% relative to the control, thyroxine (binding partner of TTR). Curation of experimental datasets from published studies resulted in 131 unique PFAS structures for modelling. The dataset was augmented with molecular descriptors and molecular docking-derived scores, followed by model optimisation and evaluation through an automated machine learning (AutoML) approach. Among the tested classification algorithms, XGBoost demonstrated the strongest overall performance, achieving a held-out test accuracy of 0.818. Structure-activity insights, molecular docking calculations and Shapley value analysis indicated an association between higher PFAS lipophilicity and Active predictions, consistent with the hydrophobic nature of the TTR binding pocket and previous experimental observations. In accordance with the FAIR principles for data and models, the curated dataset is publicly available through ChemPharos (https://db.chempharos.eu/datasets/Datasets.zul?datasetID=ds21). The final model is implemented as a user-friendly web application (https://enaloscloud.novamechanics.com/scenarios/thyrotox/).

Read the full text

Related topics

Related open access articles

Abstract and bibliographic data from PubMed (U.S. National Library of Medicine). Copyright of the abstract belongs to the original publisher or authors.