Code supporting the publication: Evaluating Molecular Representations for Predicting Cyclodextrin-PFAS Binding Energy with Machine Learning: Domain Transfer and Data Limitations

This repository contains the data and analysis code for evaluating molecular representations for predicting binding between cyclodextrins and PFAS molecules via machine learning. The project aims to explore CD-PFAS binding prediction, impacts of limited data, and most effective molecular representations for later cyclodextrin-based polymer (CDP) modeling. These models will then be used for novel CDP design for PFAS removal from water systems. Data used in this analysis comes from the OpenCycloDatabase, a consolidation of experimental binding results from published studies for cyclodextrin and guest molecule pairs, and two additional experimental studies, containing experimental binding data for cyclodextrins and PFAS molecules.

本仓库包含用于通过机器学习评估分子表征，以预测环糊精 (cyclodextrins) 与全氟和多氟烷基物质 (PFAS) 之间结合作用的数据集与分析代码。本项目旨在探索环糊精-PFAS结合预测任务、有限数据带来的影响，以及适用于后续环糊基聚合物 (CDP) 建模的最优分子表征方案。所得到的模型将可用于开发新型环糊基聚合物，以实现水系统中全氟和多氟烷基物质的去除。本分析所用数据集来源于开放环糊精数据库 (OpenCycloDatabase)——该数据库整合了已发表研究中关于环糊精与客体分子对的实验结合结果——以及另外两项实验研究，其中包含环糊精与全氟和多氟烷基物质分子的实验结合数据。