Application of the Random Forest Classification Method to Peaks Detected from Mass Spectrometric Proteomic Profiles of Cancer Patients and Controls

摘要: The random forest classification method was applied to classify samples from 76 breast cancer patients and 77 controls whose proteomic profile had been obtained using mass spectrometry. analysis consisted of two stages, the detection peaks profiles construction a rule forests. Using peak based on finding common local maxima in smoothed sample spectra, 444 were detected, reducing 365 robust found at least 7 out 10 subsets samples. Subjects classified as cases or algorithm peaks. Based prediction status out-of-bag samples, total error rate 16.3%, with sensitivity 81.6% specificity 85.7%. Measures importance each calculated identify regions spectrum influencing classification, four most important identified mz3863_13, mz2943_12, mz3193_44 mz8925_94. Combining initial provides high-performance system for data, unbiased estimates future performance.