A novel feature selection method to choose protein biomarkers for CHD

Coronary heart disease is now a severe burden on society and family in both industrialized and developing countries, but the traditional related factors can not explain the whole situations. The effect of treatment is very limited. Determination of treatment based on the symptoms and signs in Traditional Chinese Medicine (TCM) have certain curative effect. But the objective of the syndrome diagnose is a very urgent and difficult problem in TCM. Under the foundation work we had done, the aim of this study was to search for a novel feature selection method for the chose of protein biomarkers for unstable angina (UA) with blood stasis syndrome. Plasma samples were obtained from twenty-four unstable angina patients and twelve healthy volunteers. A polyclonal antibody affinity column was used to remove the six most abundant proteins. Then, the two classes of samples were separated by Two-dimensional difference gel electrophoresis (2D-DIGE). The differentially expressed protein spots were selected and identified with matrix-assisted laser desorption /ionization time-of-flight mass spectrometry (MALDI-TOF-MS) or MS-MS. Using multi-layer perceptron, we choose the protein biomarkers for UA with blood stasis syndrome. By using a back elimination method integrated with classification methods, it is found that the optimal number of proteins is six, i.e., ¿1-Antitrypsin 2, HP cleaved ß, ¿1-Antitrypsin, ApoA-I, DBP4, Fg ß 2. The proteins are biomarkers to separate UA with blood stasis syndrome from health control. The UA with blood stasis syndrome diagnosis accuracy could reach 96%. The obtained patterns are validated by 3-fold cross validation. The multi-layer perceptron may be a useful novel feature selection method for the choosing of protein biomarkers for CHD.