High-throughput screening of protein interactions with per- and polyfluoroalkyl substances (PFAS) used in photolithography.

Per- and polyfluoroalkyl substances (PFAS) are synthetic chemicals used extensively across industries, including semiconductor manufacturing. Semiconductors are ubiquitous, and there is increasing global demand for semiconductors, e.g., for advanced technologies and the automotive industry. Despite their extensive use, the toxicity and bioaccumulation potential of PFAS used in photolithography, a critical process in semiconductor manufacturing, remain poorly understood. Moreover, most lack experimental data and standards for testing. Here, we identified 96 photolithography-relevant PFAS and developed a computational framework to evaluate their potential hazards through protein binding. By integrating molecular dynamics (MD) and docking, we predicted the binding affinities and positions of PFAS to five proteins-liver fatty acid binding protein (LFABP), serum albumin (SA), peroxisome proliferator-activated receptors α and γ (PPARα and PPARγ), and transthyretin (TTR). These proteins were chosen as their binding with PFAS has been linked to PFAS bioaccumulation and to hepatic, reproductive, developmental, and endocrine disruption. Comparisons with empirical data demonstrated our approach balances simulation speed and robustness, better estimating absolute and relative binding affinities than docking alone. PFAS-protein binding affinities were generally positively associated with fluorinated chain length and the presence of aromatic rings, but limited by the protein binding pocket dimensions. Notably, we identified 22 PFAS with stronger predicted binding than perfluorooctane sulfonic acid (PFOS), a known hazardous PFAS, to at least one target protein, suggesting the potential for toxicological concern. By enabling proactive evaluation of PFAS that are unavailable for experimental testing, this work contributes to safeguarding environmental and human health amidst rising semiconductor demands.