Data-Driven Discovery of Linear Molecular Probes with Optimal Selective Affinity for PFAS in Water

• DOI: 10.1021/acs.jced.3c00404
• Publication Date: 2023-11-02T14:04:16Z
• Authors: Siva Dasetty; Maximilian Topel; Yifeng Oliver Tang; Yuqin Wang; Eric Jonas; Seth B. Darling; Junhong Chen; Andrew L. Ferguson
• Keywords: Chemical space; Molecular Recognition; Sodium dodecyl sulfate

Approaches to tackle the wide and growing variety of highly persistent per- polyfluoroalkyl substances (PFAS) are pressing global need because their detrimental human health effects, such as cancer, birth defects, hormone imbalance. Sensitive, selective, easy-to-use real-time sensors monitor detect PFAS sorbents extract them critical meeting government-mandated environmental concentrations. In this work, we combine all-atom molecular dynamics simulations, enhanced sampling, deep representational learning, Bayesian optimization perform high-throughput virtual screening for sensitive selective probes. Our design space consists 3850 linear hydrocarbon chains with varying degrees halogenation without amine- phosphine-based headgroups. By employing a data-driven search process, efficiently explore optimize sensitivity perfluorooctanesulfonic acid (PFOS) prototypical analyte selectivity relative sodium dodecyl sulfate (SDS) interferent. We calculate 504 Gibbs free energies probe-analyte probe-interferent interactions identify probes PFOS association up (−ΔGPFOS) = 9.8 ± 0.2 kJ/mol selectivities SDS (−ΔΔGPFOS–SDS) 3.1 1.5 kJ/mol. A C11Br23P(CH3)2 probe containing 11 backbone brominated carbons tertiary phosphine headgroup possesses most binding constant within defined KbPFOS 177.4 12.7, semibrominated C5H11C7Br14N(CH3)2 12 amine highest KbPFOS/KbSDS 4.6 1.7. retrospective analysis our data interpretable rules reveals that hydrogenated increases by approximately 1 per C–C bond. The addition or removal halogen atoms headgroups produces nonmonotonic changes in both 2.5 This work places empirical limitations on performance range detection offers generic strategy computational promote binding.