Background: Pancreatic cancer is a highly aggressive disease with a poor prognosis, and one in which most measure of genomic instability (e.g., tumor mutation burden (TMB), fraction of genome altered (FGA)) have thus far, not proven informative in differentiating survival. Non-B DNA are alternative DNA forms that deviate from the canonical B-DNA structure with potential to increase susceptibility to mutations, leading to the development of cancer. The role of non-B in pancreatic cancer has not been fully explored. Herein, we investigate the relationship between gene mutation sites co-localized with non-B DNA motifs in terms of survival and B-DNA features. Methods: Using TCGA data, we derived a genome-wide mutation signature on 104 early-stage pancreatic patients. We introduced a metric, mutation-localized non-B burden (MLNB) by the number of non-b motifs containing mutation sites, normalized to the number of mutations and motif library size. We applied MLNB to patient-specific mutation signatures to derive an MLNB burden for non-B DNA structure motifs: G-quadruplexes (G4), Z-DNA, inverted repeats (IR), mirror repeats (MR), direct repeats (DR), and short tandem repeats (STR). We performed a cluster analysis on MLNB and compared groups based on overall survival (OS) in months (mos) using a log-rank test. Comparisons of B-DNA-derived molecular features among clusters were performed using a Kruskal-Wallis and Fisher Exact test. Results: Among the 104 early-stage pancreatic patients with a mutation signature, MLNB clustering resulted in six patient clusters that differentiated by non-B DNA structure, with DR burden the longest OS cluster (n=23, median OS=30mos), significantly (p < 0.05) differing as compared to IR (n=23, median OS=15 mos), STR (n=20, median OS=16mos), MR with lack of IR (n=14, median OS=13mos) and MR with IR (n=16, median OS=8mos). A mix of Z-DNA and G4 burden defined a cluster with shorter OS (n=22, median OS=14mos) as compared to DR, though not significant. Patients with the longest OS, MLNB-DR burden cluster had mutation signatures enriched in MAPK and Notch signaling pathways, as compared to the other clusters enriched with double-stranded break and mismatch repair (IR), hedgehog and WNT signaling (STR), and interleukin-4 signaling (MR) pathways. Among all clusters, CDKN2A deletion was most prevalent except for MLNB-DR burden cluster. No Significant differences were found between the MLNB clusters with age, race gender, KRAS and TP53 mutations, FGA, TMB, tumor purity, Bailey and Moffit subtypes. Conclusions: Our results point towards the potential use of MLNB as an emerging marker of site-structure genomic instability. Considering that triplex DNA is emerging as a marker genomic instability, and that we distinguish between two MR groups based on IR burden, our results may provide insights into various instability mechanisms associated with poor survival in pancreatic cancer.