Abstract Molecular and immunologic studies analyzing tumor samples have failed to find a robust and reliable predictive marker of responsiveness to immune checkpoints. Exosomes are circulating microvesicles that contain a subtranscriptome of their cell of origin. They are produced by tumor and immune cells and have been shown to be mediators of immune responses in cancer. To examine the role of peripheral-blood (PB) derived exosomal transcriptomic signatures, we performed microarray analysis on N=99 samples from both PB-derived exosomes and tumor biopsies derived from N=39 (N=25 responders, N=14 non-responders) patients undergoing aPD1 immunotherapy for metastatic melanoma prior to and throughout the course of treatment. We observed increased expression of adaptive immunity, innate immunity, and antigen-presentation pathways in responders versus nonresponders in both exosomal and tumor samples prior to and during the course of treatment. We further observed that PB-derived exosomal expression profiles are highly concordant with tumor-specific expression profiles; however, they are significantly enriched in genes related to both innate and adaptive immune pathways relative to tumor expression profiles, suggesting that PB-derived exosomes share both immune-derived and tumor-derived signatures. A time-series analysis of the exosomal profiles during treatment showed significant differences in chemokine and cytokine signaling time dynamics between responders and non-responders, with IL-12 and type 1 interferon signaling pathways being the most impacted. Due to these findings, we hypothesized that exosomal profiles may serve as a predictive and prognostic tool for checkpoint blockade immunotherapy success. By utilizing markers derived from differential gene expression analysis, we were able to construct a performant machine learning classifier that can predict immunotherapy success using only pre-treatment PB-derived exosomal expression signatures. In a preliminary analysis, our model is able to achieve an auROC of 0.938 when evaluated on a N=24 (N=16 responders, N=8 non-responders) cohort using leave-one-out cross-validation, suggesting that PB-derived exosomal signatures may be predictive of immunotherapy success in metastatic melanoma. Citation Format: Alvin Shi, Jessica A. Cintolo-Gonzalez, Isabel Chien, Dennis T. Frederick, Roman Alpatov, William Michaud, Deborah Plana, David Panka, Ryan Corcoran, Keith Flaherty, Ryan Sullivan, Manolis Kellis, Genevieve Boland. Exosomal transcriptomic signatures tracks and predicts response to checkpoint blockade immunotherapy [abstract]. In: Proceedings of the AACR Special Conference on Tumor Immunology and Immunotherapy; 2017 Oct 1-4; Boston, MA. Philadelphia (PA): AACR; Cancer Immunol Res 2018;6(9 Suppl):Abstract nr B25.