MS-based Statistical Analysis for Proteomic Marker Discovery

• Activity Number: 114
• Type: Contributed
• Date/Time: Monday, August 9, 2004 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Xuena Wang*+ and Wei Zhu
• Companies: Stony Brook University and SUNY, Stony Brook
• Keywords: Gaussian random field ; stepwise discriminant analysis ; K-nearest neighbor method ; cross-validation ; resampling ; multiple comparison

Abstract

With the maturation of the new Mass Spectrometry (MS) technologies, as well as the availability of a well-defined clinical database, the MS-based statistical analysis of volume-produced serum or plasma mass spectra retrieved from healthy and disease-affected individuals plays an increasingly important role in the field of proteomic research, especially the diagnosis of disease and the follow-up of treatment effect. To decrypt a disease-specific proteomic pattern is extremely challenging considering the complexity of the sample individuality and the multitude of disease stages and combinations. For a large scale screening of diseases with relative low prevalence such as the ovarian cancer, Positive Predictive Value (PPV) of 100% becomes critical. We propose a novel statistical procedure identifying the best model of biomarkers for the optimal classification between two different groups of subjects. In comparison with other approaches reported to date, this procedure is shown to be more robust and have higher discriminatory power when applying to the ovarian cancer study in the analytical platforms of both low-resolution PBS-II TOF MS and high-resolution QqTOF MS.