BACKGROUND. Mitochondrial diseases belong to the group of inborn errors of metabolism (IEM), with a prevalence of 1 in 2,000–5,000 individuals. They are the most common form of IEM, but, despite advances in next-generation sequencing technologies, almost half of the patients are left genetically undiagnosed. METHODS. We investigated a cohort of 61 patients with defined mitochondrial disease to improve diagnostics, identify biomarkers, and correlate metabolic pathways to specific disease groups. Clinical presentations were structured using human phenotype ontology terms, and mass spectrometry–based proteomics was performed on primary fibroblasts. Additionally, we integrated 6 patients carrying variants of uncertain significance (VUS) to test proteomics as a diagnostic expansion. RESULTS. Proteomic profiles from patient samples could be classified according to their biochemical and genetic characteristics, with the expression of 5 proteins (GPX4, MORF4L1, MOXD1, MSRA, and TMED9) correlating with the disease cohort, thus acting as putative biomarkers. Pathway analysis showed a deregulation of inflammatory and mitochondrial stress responses. This included the upregulation of glycosphingolipid metabolism and mitochondrial protein import, as well as the downregulation of arachidonic acid metabolism. Furthermore, we could assign pathogenicity to a VUS in MRPS23 by demonstrating the loss of associated mitochondrial ribosome subunits. CONCLUSION. We established mass spectrometry–based proteomics on patient fibroblasts as a viable and versatile tool for diagnosing patients with mitochondrial disease.