Abstract Pancreatic adenocarcinoma (PDAC) is the third leading cause of cancer mortality, reflecting the fact that only 9% of patients present with tumor localized to the primary site, so that the vast majority of patients with pancreatic cancer are unresectable at the time of diagnosis. We have created a program that spans the identification of new molecular targets by spatial transcriptomics, the identification of peptide/small molecule ligands, and the preclinical and eventual clinical evaluation of these therapies. Peptide targeted radionuclide therapy (PTRT) is a molecularly targeted treatment strategy that uses radiolabeled peptides as biological targeting vectors designed to deliver cytotoxic levels of radiation dose precisely to cancer cells that overexpress specific receptors. Peptides with high receptor affinity are conjugated with a chelator for diagnostic imaging using positron emission tomography (PET) for patient staging, selection of suitable candidates for PTRT, and subsequent treatment monitoring. In turn, peptides are then stably radiolabeled with the intermediate-energy β-emitter lutetium-177 (177Lu). Peptides are superior for delivery due to the ease of synthesis, comparable potential affinity and specificity, improved pharmacokinetic profiles, and low immunogenicity. We now have 40 spatially sequenced pancreatic cancer samples, each with 4000 locations sequenced at a depth of 18,000 genes. Across all of these samples, we have identified Claudin-4 (CLDN4) as a particularly promising target; CLDN4 is the top target from our analysis as it 1) is a tight junction protein that is not accessible in normal tissue, and 2) is overexpressed (~16x) in all patients and in 97% of the cancer cluster spatial regions we have sequenced. Peptides have been identified with nanomolar affinity for CLDN4 and we have demonstrated specific accumulation of ~20%IA/cc in tumor and peritoneal metastases in a PDAC model. Our spatial analysis of these samples indicates that if we can deliver a peptide carrying a beta-emitter to all CLDN4 expressing cells, all cancer cells could receive a therapeutic dose of radiation. Further, we have developed the first computational method to generate cyclic peptides to specific binding domains and have now generated peptides against the tight junction region of CLDN4. Clostridium perfringens enterotoxin (Cpe) is a natural ligand to CLDN4, and we evaluate mutated peptide fragments from this ligand and cyclic peptides designed by new machine learning tools. Our specific aims are to: Aim 1) Assess and optimize claudin-4 peptides in vitro. Aim 2) Evaluate 64Cu and 68Ga-labeled claudin-4 peptides in cell culture and mouse models of pancreatic cancer Aim 3) Evaluate claudin-4 peptides labeled with lutetium-177 in 3A) cell culture, 3B) xenograft, transgenic and PDX mouse models and 3C) combination therapy with immunotherapy.