Background: Metastatic colorectal cancer (CRC) patients who have become resistant to standard-of-care (SOC) treatments are often put on targeted therapies, including experimental drugs still in clinical trials. However, patients with targeted mutations still often do not respond to the targeted therapies alone, hence it remains a critically unmet need to explore potential combinatorial regimens that will enhance the efficacy. Methods: We performed a high-throughput screen using MicroOrganoSpheres (MOS) derived from metastatic refractory patients to correlate with clinical outcomes and explore alternative combinations that might benefit the patients. Biopsies from metastatic CRC patients who became resistant to SOC and were about to receive targeted therapies in clinical trials were molecularly profiled and implanted into immunodeficient mice to generate patient-derived xenografts (PDX). Results: MOS were derived and established from PDX using droplet-based microfluidics for high-throughput 3D screening. These MOS-based refractory patient avatars contain either targetable mutations including KRAS G12C and BRAF 600E or no such mutations. Response of the MOS to dose-titrated single and combination drug treatments including the available targeted therapies were measured by a fully automated robotic dispensing and imaging pipeline and quantified by AI-based live imaging analytical algorithm to generate robust drug response curves, which were used to derive IC50 and EC50. We tested both FDA-approved and experimental targeted therapies as single agent or in combination in 5-dose titrations. Longitudinal tracking of each MOS with Live/dead fluorescence dye signals was obtained for 5 days. MOS drug response curves and IC50s show correlation with clinical outcomes. This comprehensive dataset enabled us to assess the potential benefit of different combinations, help deconvolute the efficacy and synergy of individual drugs, and provide insight into patients who fail to response to targeted therapies despite possessing the targeted mutations. As the MOS diagnostic assay can be completed from a patient biopsy within 10 days and is currently in a registered, multi-site clinical trial (ClinicalTrials.gov # NCT05189171), this study provides proof-of-principle for a potential clinical MOS-based assay to guide refractory patients to the optimal combination regimen containing targeted therapy. Conclusions: Refractory patients may benefit from a MOS-based high-throughput screen assay to select the optimal combination regimen containing targeted therapy.