Abstract Numerous next-generation sequencing studies have provided an overview pancreatic ductal adenocarcinoma (PDAC)’s genomic evolution. But how its genome evolves through treatment and metastasis has not been extensively studied due to lack of samples and technical limitation of bulk sequencing on low-tumor purity samples. Herein we applied targeted single-nucleus DNA sequencing (snDNA-seq) to a cohort of 18 patients, each with >= 2 multiregional/longitudinal samples, totaling 65 samples. These included 5 early-stage resections, 11 stage IV autopsies and 2 biopsy-derived organoids taken pre- and post-treatment. We also developed a set of computational methods for this new datatype to delineate genomic evolution at high resolution. Despite higher sensitivity than bulk, snDNA-seq did not uncover significant subclonal single-nucleotide variations (SNVs) on our panel targeting 253 select genes for PDAC. But it detected a higher frequency of CDKN2A (72% against 30%) and SMAD4 (56% against 32%) alterations, mostly attributed to chromosomal deletions. Many were subclonal or only affected a few hundred base-pairs of the gene, making them elusive to bulk sequencing. Except for one case driven by BRCA2 mutations rather than the canonical KRAS oncogene, other patients had mostly linear phylogenies with more SNVs occurring before the most recent common ancestor (MRCA) of clones, for early/late-stage and longitudinal cases alike. This early fixation of drivers and rapid clonal sweep afterwards likely correlates with the particular aggressiveness of PDAC. Metastatic clones almost always fell towards the end of phylogenies and were shared among spatially separated sites, indicating late metastatic dissemination in molecular evolution time. Metastases to the liver and diaphragm appeared to be more genetically evolved than other sites, likely due to harsher selective pressure cast by the metastatic routes/distal microenvironments. 16 of 18 PDACs were observed to converge towards tumor cell-intrinsic TGF-β unresponsiveness by mutating various components of the pathway. Such convergent evolution usually happened at the primary site, indicating a strong selective advantage of the phenotype in the desmoplastic, nutrient-poor pancreas microenvironment. Continuous evolution was seen through treatment and metastasis, driven by seemingly random genome-scale copy number variations (CNVs) and focal amplification/deletions to genes such as KRAS, CDKN2A, SMAD4, MYC. These insights on PDAC’s genomic evolution inform more precision medicine efforts to come. Early fixation of driver SNVs sculpts a largely homogeneous disease that could be uniformly targeted. But special care should be taken to obviate resistance mechanisms conferred by continuously evolving CNV events- compared to SNVs, these mechanisms provide faster remodeling of the genome, and thus faster generation of new phenotypes and adaptation. Citation Format: Haochen Zhang, Palash Sashittal, Elias-Ramzey Karnoub, Akhil Jakatdar, Shigeaki Umeda, Nicolas Lecomte, Jungeui Hong, Katelyn Mullen, Akimasa Hayashi, Caitlin A. McIntyre, Benjamin J. Raphael, Christine A. Iacobuzio-Donahue. A refined view of pancreatic cancer genomic evolution through single-nucleus DNA sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6927.