1507P Revolutionizing clinical trials: Harnessing artificial intelligence for optimized phase III studies in pancreatic cancer

factors.Feature selection for S-vs.L-DFS prediction (n 80 Pts; PO: n 42, A: n 38) revealed a panel of 8 proteins for the PO group (Nave Bayes; CSF3, LGALS3, GZMA, CCL1, IL20, IL3, CCL7, PTX3), 6 proteins for the A group (J48; IL1A, IL1B, IL3, IL34, MIF, CCL19), and 7 proteins for the combined (c)rPDAC group (MLP; CCL21, CX3CL1, IL16, IL20, IL7, PTX3, TSLP).ML successfully predicted S-/L-DFS for the PO (via MLP), crPDAC (via MLP), and A (via J48) group (ROC-AUC (training) > 0.9, ROC-AUC (validation) > 0.9, respectively).Conclusions: In our study, we have identified a favorable TiME for L-DFS rPDAC Pts under PO Gem-nabPac CTX.We further present minimal immune-related liquid biomarker signatures based on mELISA and ML for early prediction of Gem-nabPac CTX success in PO and A treatment settings, or for rPDAC Pts in general.