Machine learning assisted photothermal conversion efficiency prediction of anticancer photothermal agents

Abstract

Photothermal therapy (PTT) is a minimally invasive and promisingly effective strategy for thermal ablation of tumors. There is an urgent need for the development of ideal organic photothermal agents (PTAs) with high photothermal conversion efficiency (PCE). Machine learning (ML)-assisted predictions of PCE could offer an efficient way for early screening of PTAs. Herein, 44 organic PTAs were collected from the literature as a dataset to establish a best-performed regression model by comparing different ML methods, in which R2, Pears, and RMSE were 0.761, 0.913, and 0.058, respectively. Then, the reliability of the model was further verified by predicting two newly designed PTAs. The double bond of tetraphenylethylene (TPE) was found to be an important substructure to enhance PCE by the Shapley additive explanations method. The results show that ML can provide a valuable tool for predicting PCE of PTAs, thus promoting the development of photothermal therapy for cancer.