A Machine Learning Approach for Predicting Defluorination of Per- and Polyfluoroalkyl Substances (PFAS) for Their Efficient Treatment and Removal

Abstract

We present the first application of machine learning on per-and polyfluoroalkyl substances (PFAS) for predicting and rationalizing carbon−fluorine (C−F) bond dissociation energies to aid in their efficient treatment and removal. Using a variety of machine learning algorithms (including Random Forest, Least Absolute Shrinkage and Selection Operator Regression, and Feed-forward Neural Networks), we were able to obtain extremely accurate predictions for C−F bond dissociation energies (with deviations less than 0.70 kcal/mol) that are within chemical accuracy of the PFAS reference data. In addition, we show that our machine learning approach is extremely efficient, requiring less than 10 min to train the data and less than a second to predict the C−F bond dissociation energy of a new compound. Most importantly, our approach only needs knowledge of the simple chemical connectivity in a PFAS structure to yield reliable resultswithout recourse to a computationally expensive quantum mechanical calculation or a threedimensional structure. Finally, we present an unsupervised machine learning algorithm that can automatically classify and rationalize chemical trends in PFAS structures that would otherwise have been difficult to humanly visualize or process manually. Collectively, these studies (1) comprise the first application of machine learning techniques for PFAS structures to predict/rationalize C−F bond dissociation energies and (2) show immense promise for assisting experimentalists in the targeted defluorination of specific bonds in PFAS structures (or other unknown environmental contaminants) of increasing complexity.

Real-time degradation dynamics of hydrated per- and polyfluoroalkyl substances (PFASs) in the presence of excess electrons

“…This facile dissociation of secondary C-F bonds is due to their lower bond dissociation energy compared to the primary C-F bonds, as noted in a few earlier studies. 23,24 Contrary to the addition of a single electron, we find that the spin-density can evolve in two different ways with the addition of multiple electrons, as shown in the middle and bottom panels of figure 3. We refer to these different regimes as a trans-type (middle panel) or cis-type elimination (bottom panel) based on how the fluorine atoms have dissociated from the PFAS backbone structure.…”

A database framework for rapid screening of structure-function relationships in PFAS chemistry

“…PFAS-Map can also be coupled with dissociation data to study the structure-persistence relationship of PFASs. Figure 9 shows the mean C-F bond dissociation energy (the average of all C-F bonds’ dissociation energy in a molecule) calculated based on Raza et al .’s work on machine learning prediction of PFAS defluorination 15 . The PFAS map highlights the trend that the mean dissociation energy generally decreases as the length of perfluoroalkyl chain increases, and also that the mean dissociation energy for aromatic PFASs is significantly higher than those aliphatic PFASs with a similar number of carbons.…”

“…

Fig. 9 PFAS-Map showing the predicted mean C-F bond dissociation energy from the Raza et al .’s work “A Machine Learning Approach for Predicting Defluorination of Per and Polyfluoroalkyl Substances (PFAS) for Their Efficient Treatment and Removal” 15 . An interactive version of this figure is provided in figshare File 1.

…”

Deep learning-based prediction of future growth potential of technologies

“…The t-SNE algorithm visualizes high-dimensional data in two dimensions, and the mapped data points (i.e., technology clusters) can be clustered based on their similarities. Therefore, t-SNE algorithm can be a useful tool for automatic visualization and clustering of large data sets [ 36 ]. Like the similarity-based map, each technology cluster in CWTS has a different color depending on the field.…”