Prediction of PFOA and PFOS Toxicity through Log P and Number of Carbon with CompTox and Machine Learning Tools

Perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS) are two major groups of PFAS will be subjected to the Maximal Contamination Concentration (MCL) of 4 ng/l in drinking water to be implemented by the U.S. EPA by 2025. How to accurately predict toxicity of PFAS with varied carbon chain length is important for treatment and sequential removal from drinking water. This study presents Quantitative Structure and Activity Relationship (QSAR) models developed through both linear regression and two order regression. Log P is compiled from reference and carbon content is counted as the molecule represented. Bioconcentration potential is predicted from CompTox.The results suggest that as log P and carbon content increase, the bioconcentration potential of PFCAs also increases. In other words, larger PFCA molecules tend to be easier to bioaccumulate in living organisms. This finding is crucial because bioconcentration refers to the accumulation of substances from water directly into living organisms through the process of passive diffusion across cell membranes. On the other hand, 96-hour fathead minnow LC50 has an inverse relationship, with higher LC50 values associated with lower log P and fewer carbons. The varying R-squared values across methods indicate differing degrees of correlation, underscoring the impact of compound structure on aquatic toxicity. Similarly, for oral rat LD50 and 48-hour D. magna LC50, the R-squared values reflect moderate to strong correlations with log P and the number of carbons. As the log P and carbon content decrease,the toxicity expressed in LC50 or LD50 increases. This relationship underscores the role of chemical properties in influencing the toxicity of PFCAs across different organisms and exposure routes. For instance, the negative correlation between log P and aquatic toxicity (96-hour fathead minnow LC50 and 48-hour D. magna LC50) suggests that compounds with higher hydrophobicity (higher log P) and more carbons may exhibit lower acute toxicity to aquatic organisms.