Biomarker development is a major research theme in cancer proteomics. Cancer is a genetically and clinically diverse disease, and biomarkers for risk stratification therapy are urgently required. A considerable number of biomarker candidates have been discovered by proteomics, and over the last decade proteomics modalities have been developed to identify promising candidates. Validation studies involving hundreds of samples in independent cohorts is the next challenge to prove the clinical utility of any discovered biomarker candidates. Here, we review our efforts directed toward tissue biomarker development using a proteomics approach. With the aim of developing a prognostic biomarker for gastrointestinal stromal tumor (GIST), we examined the protein expression profiles of primary tumors from 17 GIST patients with different risks of recurrence and prognosis after surgery. Through a comparative study using two-dimensional difference gel electrophoresis and mass spectrometry, we found that overexpression of pfetin was specific to GIST patients with a low risk of metastasis and a favorable prognosis after surgery. Using immunohistochemistry, we examined pfetin expression in 422 additional cases of GIST at four hospitals, and confirmed that GIST patients with pfetin-positive primary tumors had a significantly favorable prognosis in all four cohorts. Moreover, the other research group independently validated the prognostic significance of pfetin in 62 cases of GIST at two hospitals. Pfetin was found to be an independent prognostic factor with significant prognostic utility in all risk classification groups, which are based on tumor size and mitosis status. In addition to pfetin, we also identified DDX39 as a biomarker of unfavorable prognosis using a proteomics approach, and KCTD10 as a marker of favorable prognosis using a knowledge-based approach. Our experience demonstrates the utility of proteomics for biomarker discovery, and the possible clinical application of pfetin for risk stratification therapy in GIST.