AIM: A discrimination analysis has been explored for the probabilistic classification of healthy versus ovarian cancer serum samples using proteomics data from mass spectrometry (MS). METHODS: The method employs data normalization, clustering, and a linear discriminant analysis on surface-enhanced laser desorption ionization (SELDI) time-of-flight MS data. The probabilistic classification method computes the optimal linear discriminant using the complex human blood serum SELDI spectra. Cross-validation and training/testing data-split experiments are conducted to verify the optimal discriminant and demonstrate the accuracy and robustness of the method. RESULTS: The cluster discrimination method achieves excellent performance. The sensitivity, specificity, and positive predictive values are above 97% on ovarian cancer. The protein fraction peaks, which significantly contribute to the classification, can be available from the analysis process. CONCLUSION: The discrimination analysis helps the molecular identities of differentially expressed proteins and peptides between the healthy and ovarian patients.
AIM: A discrimination analysis has been explored for the probabilistic classification of healthy versus ovarian cancer serum samples using proteomics data from mass spectrometry (MS). METHODS: The method employs data normalization, clustering, and a linear discriminant analysis on surface-enhanced laser desorption ionization (SELDI) time-of-flight MS data. The probabilistic classification method computes the optimal linear discriminant using the complex human blood serum SELDI spectra. Cross-validation and training/testing data-split experiments are conducted to verify the optimal discriminant and demonstrate the accuracy and robustness of the method. RESULTS: The cluster discrimination method achieves excellent performance. The sensitivity, specificity, and positive predictive values are above 97% on ovarian cancer. The protein fraction peaks, which significantly contribute to the classification, can be available from the analysis process. CONCLUSION: The discrimination analysis helps the molecular identities of differentially expressed proteins and peptides between the healthy and ovarianpatients.
Authors: Nancy Leymarie; Mark E McComb; Hicham Naimy; Gregory O Staples; Joseph Zaia Journal: Int J Mass Spectrom Date: 2011-07-23 Impact factor: 1.986