Background: Despite recent successes with immune checkpoint blockade (ICB) in melanoma, the prognosis for most patients remains dire. Whereas small fraction of patients are able to achieve disease control, most do not respond or are limited by immune-related toxicities. Robust non-invasive predictors of ICB response have the potential to guide clinical decision and alter management of patients, however, no such predictors currently exist. Methods: We applied a highly-multiplex Proximity Extension Assay to simultaneously detect > 1000 proteins in the plasma of anti-PD-1 treated melanoma patients. Our cohort comprised 116 patients, 66 responders (R) and 50 non-responders (NR). Additional 65 patients comprised a validation cohort with 30 R and 35 NR, and included 50 patients who developed treatment-related toxicities. Plasma samples were collected at baseline, 6-weeks and 6-months after starting the treatment. A subset of patients had single-cell RNA-seq performed on tumor tissue. Group differences and treatment effects were evaluated by linear model with maximum likelihood estimation for model parameters and Benjamini and Hochberg multiple hypothesis correction. Results: At baseline, 6 significantly differentially expressed (DE) proteins were identified between R and NR. Elevated expression of ST2 and IL-6, two key immunoregulatory proteins were found in NR. At 6-weeks, more dynamic changes occurred and 79 significantly DE proteins were identified between R and NR, including proteins implicated in primary or acquired resistance as IL-8, MIA, TNFR1 and potential novel targets as MCP-4/CCL13, ICOSLG and VEGF. Proteomic changes identified at baseline and 6-weeks were more profound at 6-months, and moreover 238 proteins were confirmed significant between R and NR. Importantly, we were able to leverage these differences to build classifiers of R and NR subsets. We compared mRNA expression of DE proteins within the tumor microenvironment by leveraging scRNAseq data from a subset of these patients. Enriched expression of these genes was uncovered in certain myeloid and exhausted T cell subsets, thus shedding insight into the potential role of these cell subsets in ICB response. Conclusions: Plasma proteomic profiling of anti-PD1 treated patients identified important tumor and immune changes associated with response. Non-invasive means discovery of circulatory protein biomarkers may predict sensitivity to immunotherapy and uncover biological insights underlying primary resistance.