Assessment of Drug Susceptibility for Patient-Derived Tumor Models through Lactate Biosensing and Machine Learning.

Abstract

A patient-derived tumor model (PDM) is a practical tool to rapidly screen chemotherapeutics for individual patients. The evaluation method of cell viability directly determines the application of PDMs for drug susceptibility testing. As one of the metabolites of "glycosis", the lactate content was used to evaluate cell viability, but these assays were not specific for tumor cells. Based on the "Warburg effect", wherein tumor cells preferentially rely on "aerobic glycolysis" to produce lactate instead of pyruvate in "anaerobic glycolysis" of normal cells, we reported a gold lactate sensor (GLS) to estimate the cell viability of PDMs in drug susceptibility testing. It demonstrated high consistency between the GLS and commercial cell viability assay. Unlike either imaging or cell viability assay, the GLS characterizes the cell viability, enables dynamic monitoring, and distinguishes tumor cells from other cells. Moreover, machine learning (ML) was employed to perform a multi-index assessment for drug susceptibility of PDMs, which proved to be accurate and practical for clinical application. Therefore, the GLS provides an ideal drug susceptibility testing tool for individualized medicine.