Abstract To maximize cancer patients’ survival rate post-therapy, in vitro immortal cancer cell models and newly developed patient-derived organoids are widely used to study the role of tumor metabolism reprogramming in tumor growth and survival under therapeutics stresses. Although conducting longitudinal metabolic measurements on the same tumor sample during a course of therapy is critical for therapeutic studies, there are surprisingly few techniques that can provide a systems-level view of tumor metabolism on in vitro cancer models or organoids non-destructively. Several metabolic tools, such as Seahorse Assay and Metabolomics, provide standardized metabolic measurements but often require destructive sample preparation. Relying on the non-invasive nature of optical technique, this proposal seeks to fill the critical technical gap by developing an optical spectroscopic assay that will enable non-destructive high-throughput metabolism measurement on in vitro cancer models and organoids for cancer research. Specifically, we will develop a novel multi-channel fluorescence spectroscopic assay and a machine learning de-convolution algorithm to quantify the key metabolic parameters of in vitro cancer models (Aim 1). As there is a significant unmet clinical need for breast cancer (BC) radiotherapy (RT) sensitivity evaluation prior to treatment, we will demonstrate our non-destructive assay for early prediction of BC radiation responses within the decision-making window via longitudinal metabolic characterization of patient-derived organoids under radiation stresses (Aim 2). Our technology fills an important gap that exists between Seahorse Assay (in vitro cells) and Metabolomics (in vitro cells and ex vivo tissue) by providing a novel approach for non-destructive metabolism measurement on in vitro cancer models and patient-derived organoids. Our innovative RT sensitivity prediction model will directly impact BC patients by providing a novel paradigm for patients’ RT sensitivity prediction during the decision-making window. Once we demonstrate the proof-of-concept of our optical technique and the RT sensitivity prediction model, we will move our study to a large-scale trail in clinics with a goal of providing individualized RT for BC patients in our future R01 plan.