Hepatotoxicity assessment investigations on PFASs targeting L-FABP using binding affinity data and machine learning-based QSAR model

Abstract

Per- and polyfluoroalkyl substances (PFASs) are persistent organic pollutants that have been detected in various environmental media and human serum, but their safety assessment remains challenging. PFASs may accumulate in liver tissues and cause hepatotoxicity by binding to liver fatty acid binding protein (L-FABP). Therefore, evaluating the binding affinity of PFASs to L-FABP is crucial in assessing the potential hepatotoxic effects. In this study, two binding sites of L-FABP were evaluated, results suggested that the outer site possessed high affinity to polyfluoroalkyl sulfates and the inner site preferred perfluoroalkyl sulfonamides, overall, the inner site of L-FABP was more sensitive to PFASs. The binding affinity data of PFASs to L-FABP were used as training set to develop a machine learning model-based quantitative structure-activity relationship (QSAR) for efficient prediction of potentially hazardous PFASs. Further Bayesian Kernel Machine Regression (BKMR) model disclosed flexibility as the determinant molecular property on PFASs-induced hepatotoxicity. It can influence affinity of PFASs to target protein through affecting binding conformations directly (individual effect) as well as integrating with other molecular properties (joint effect). Our present work provided more understanding on hepatotoxicity of PFASs, which could be significative in hepatotoxicity gradation, administration guidance, and safer alternatives development of PFASs.

Keywords

• PFASs
• L-FABP
• Binding affinity
• Machine learning
• Bayesian Kernel Machine Regression