TMET-08. METABOLOMIC CHANGES IN GLIOBLASTOMA PATIENTS UNDERGOING CONCURRENT CHEMORADIATION THERAPY

Abstract OBJECTIVE To illustrate changes of untargeted metabolomics in patients with glioblastoma IDH wildtype undergoing concurrent radiation therapy (RT) temozolomide (TMZ). This study implemented machine learning (ML) algorithms to predict treatment phase: pre-surgery, post-surgery, pre-radiation, and post based on data. METHODS Thirty-six (18 methylguanine methyltransferase [MGMT] methylated, 16 MGMT unmethylated, 2 status unknown) were enrolled into this study. Serum samples obtained from the same day before surgery, days after starting their chemoradiation, concluding phase treatment. Blood via antecubital phlebotomy without regard for time day, diet, or fasting status. Untargeted by GC-TOF mass spectrometry compared. The proposed ML models analyzed 105 36 utilizing 157 structurally identified blood metabolites. Multinomial Logistic Regression (MLR) GradientBoostingClassifier (GB Classifier) used classify patient detected classification performance these was evaluated using metrics AUC scores. RESULTS Post radiation; significant increase following metabolites: glycine, serine, threonine, oxoproline, 6-deoxyglucose, gluconic acid, glycerol-alpha-phosphate, ethanolamine, propyleneglycol, triethanolamine, xylitol, fumaric acid noted while decrease 3-aminopiperidine 2,6-dione radiation. MLR produced 78% accuracy, 75% precision, = 0.89, GB Classifier 77% precision 0.91. Finally, we presented a pattern metabolites per clinical stage pairwise correlations. CONCLUSIONS represent first serum metabolic signature associated RT glioblastoma. results correlations showed that have potential potentially enable evaluate response relatively small cohort.