Genomic alterations drive the tumorigenesis of pancreatic cancer (PC). However, alone they do not explain its numerous phenotypes. Exploring the epigenetic landscapes of PC delivers a more insightful picture and might reveal excellent targeted therapies that could improve patient survival. PC subtyping based on histological features reflects its morphological diversity and correlates with clinical outcomes. Here we used a label-free multiplexed molecular imaging to recognize PC epigenetic modifications spatially, consequently, DNA and histone methylation (at lysine and arginine) and histone acetylation (at lysine) were investigated. To complete the picture, B-to-Z-DNA conformational change was assessed. We utilized convolutional neural networks and other machine learning approaches to analyze and semi-quantify the relative variability of epigenome among the six most common PC histological subtypes. We found foamy-glands (FG) and squamous-differentiated (SD) presenting oppositely to others and more alike the benign controls. They consistently expressed higher global levels of epigenetic modifications and higher Z-DNA ratios. Overall, our results suggest variable efficacy of targeting epigenetic regulators in histologically distinct PC subtypes.