SUMMARY Glioblastoma is the most common primary brain tumor with substantial genomic, molecular and phenotypic heterogeneity, but uniformly dismal outcomes despite the current standard treatment of concurrent temozolomide chemo-radiotherapy (CRT). Given the pace of disease recurrence and the challenges associated with obtaining tumor tissue, there is an unmet clinical need for the real-time, noninvasive assessment of GBM responsiveness to CRT. As demonstrated previously, most cancers shed tumor-derived fragmented DNA into biofluids, including plasma and cerebrospinal fluid (CSF), and these cell-free molecules can be quantified as a measure of disease burden. The approach, named “liquid biopsy”, has recently emerged as a breakthrough diagnostic and monitoring tool for diseases such as cancer, with the added benefit of being minimally invasive. Through the sampling and analysis of biofluids, a number of promising glioma biomarkers, derived from tumor-derived DNA in plasma (ctDNA) and CSF (CSF-tDNA) (together called rtDNA), have been reported as diagnostic strategies for gliomas. Meanwhile, numerous previous studies have demonstrated that protein-based amide proton transfer (APT) MRI can accurately identify tumor burden and genetic markers (such as IDH, MGMT status) in gliomas. The goals of this proposal are to combine ctDNA and CSF-tDNA with APT MRI to resolve the diagnostic challenges associated with discriminating treatment effect from tumor progression and to develop an efficient and reliable deep-learning framework for post-treatment monitoring. We propose the following specific aims to be performed: (1) correlate ctDNA and CSF-tDNA levels with protein-based APT MRI characteristics when monitoring GBM treated with CRT; (2) determine the accuracy of combined rtDNA/APT indices in identifying GBM recurrence; and (3) develop a transformer pipeline using rtDNA and mpMRI to assess GBM prognosis. The success of this aim will help to understand the dynamic patterns of rtDNA/APT throughout the treatment course for individuals with GBM. If our rtDNA/APT investigation is successful, the results would dramatically improve the care of patients treated with CRT and spare many patients from undergoing surgery for diagnostic purposes.