Background: Tumor mutational burden (TMB), PDL1 status, microsatellite status (MS), tumor infiltrating lymphocyte, and the microbiome profile are commonly used biomarkers to predict which patients (pts) may benefit from immunotherapy (IO). However, they are imperfect and cannot be used to predict complete response (CR). Next Generation Sequencing (NGS) from Foundation Medicine has identified many gene mutations. We aim to identify potential biomarkers in pts who achieved CR with immunotherapies as opposed to in those who had upfront progression of disease (PD). Methods: We conducted a retrospective analysis of pts with metastatic solid tumors at our institution that received IO after NGS testing. Pts who achieved CR or PD per imaging studies and clinical settings were included. Gene alterations through NGS were identified in each group. Results: Eighty-one pts received IO who had NGS sent prior to treatment. Pembrolizumab, ipilimumab, nivolumab, durvalumab and atezolizumab were used as single agent or in combination with another IO or chemotherapy. 19 pts achieved CR, 18 pts had PD. In the CR group, melanoma and lung adenocarcinoma were most common tumor types (Table). 15 pts (78.9%) received IO as 1^st line treatment. Most common gene mutations in CR group were ROS1 (52.6%), LRP1B (52.6%), GRIN2A (47.4%), and EPHA7 (42.1%) in 19 patients. 18 pts with various tumor types (Table) had PD. Most of IOs were given as 2^nd or 3^rd line setting in NR group. Most common gene alterations were MLL2 (33.3%), CDKN2A/B (27.8%), NTRK1 (27.8%), ARID1B (22.2%) in 18 patients. Conclusions: The cohort of patients who achieved CR with IO had the most common gene mutations of ROS1, LRP1B, GRIN2A, and EPHA7 which can potentially serve as a biomarker to predict response to IO. Increased frequencies of MLL2, CDKN2A/B, NTRK1, ARID1B mutations were found in patients with PD. The gene mutations in CR and PD groups are mutually exclusive. Further explorations at tumor environment, RNA and proteins levels are needed to identify combined potential biomarkers to predict IO response, followed by validation in the prospective study. The limitation is that the study is retrospective.