OBJECTIVES/GOALS: Single-cell and spatial transcriptomics have revealed high heterogeneity in the tumor and microenvironment. Identifying populations of cells that impact a patient’s prognosis is an important research goal, so researchers can generate hypotheses and clinicians can provide targeted treatment. METHODS/STUDY POPULATION: DEGAS uses deep-transfer-learning to identify patterns between patient tumor RNA-seq and clinical outcomes and map these associations on to higher-resolution data like spatial and single-cell transcriptomics. We apply DEGAS to prostate and pancreatic cancer spatial transcriptomics samples, as well as one normal sample of prostate tissue. We used the TCGA prostate cancer cohort to with the accompanying survival information and publicly accessible prostate cancer ST data from 10X Genomics to predict survival associations in the ST slides derived from the TCGA patients. Based on these survival associations, we identify higher risk subsections of ST slides which can be further studied. RESULTS/ANTICIPATED RESULTS: We were able to validate our method by comparing it to Scissor and were able to show that the number of high-risk regions in prostate cancer slides increased with the stage of disease. Furthermore, we identify transcriptomic signatures enriched for ontology terms associated with growth regulation and apoptosis, inflammation, immune signaling, and autophagy in histologically normal prostate tissues and adjacent normal pancreatic cancer tissues that were identified as high-risk by DEGAS. The regions highlighted by DEGAS could reflect transcriptional precursors to intraepithelial neoplasia–a well-recognized premalignant morphological change in glandular epithelium. DISCUSSION/SIGNIFICANCE: Identifying biomarkers of tissue stress that precede morphologic diagnosis of high-grade pre-malignant lesions by a pathologist may help triage patients at high risk for future development of cancer, or aid in better understanding whether histologically normal pre-malignant tissues at tumor margins contribute to recurrence.