Proteomics-based Biomarker Discovery: The Need for Less Complicated Methodologies

Introduction: Although proteomic technologies hold much promise, the discovery of meaningful tumor biomarkers using this methodology is challenging. One issue is the size of the databases, and in spite of utilizing costly bioinformatics software, tumor marker detection is time-consuming. While engaged in a mass-spectrometry based project for biomarker discovery in head and neck cancer, we found that combining methodologies gave quicker and more meaningful results. Methods: Archived, formalin-fixed, paraffin-embedded tissues from twelve cases of head and neck squamous cell carcinoma, and a matching set of control (normal) tissues formed the basis for the study. Both commercially available (Expression Pathology [EP]), and in-house reagents were used for protein extractions. The EP digestion protocol with Protease Max released peptides from a one-hour trypsin digestion in both instances. A Thermo Deca LCQ was used for mass spectrometric analysis. Results: Data from twenty-four paired samples by each method was entered using spectral quantitative counts obtained from the Mascot/Scaffold/X-!Tandem program. Instead of relying on pathway analysis software alone for biomarker analysis, we opted for a two-column MudPIT format. We first derived P-values (Fisher's Exact Test), and the ‘normal’ and ‘tumor’ columns were sorted by ascending values. This brought the ‘zero’ values to the top of the ‘normal’ column, enabling potential biomarker proteins to be expressed only on the ‘tumor’ side. By further analysis utilizing bioinformatics software (Ingenuity Pathway Analysis – IPA), we could narrow the search down to less than ten relevant candidate proteins that were common to both analyses. Those with high enough P-values, and good interactions via IPA were selected for immunohistochemical validation as tumor biomarkers. Conclusions: Using the simpler MudPIT analytical approach in conjunction with IPA, faster discovery times of biomarker candidates may be possible. Further validation awaits application of these techniques on a larger sample of cases.