Time filter

Source Type

Shijiazhuang, China

Chen J.,Beijing University of Chinese Medicine | Jia Z.,Academy of Hebei Province | Wu X.,Academy of Hebei Province | Yuan G.,Academy of Hebei Province | And 4 more authors.
Journal of Biological Systems | Year: 2010

Hyperlipidemia (HL) and unstable angina (UA) are two sequential diseases that cause more and more morbidity and mortality world-wide. Biomarkers selection in the level of physical and chemical specifications (PCS) plays a key role in understanding the pathology of both diseases. Neuro-Endocrine-Immune (NEI) system is a preferable pathway to investigate the interaction network of related PCS in the context of HL and UA. Data mining approaches are a kind of advanced statistical methods to unravel the "secret" of interaction network of PCS in both diseases. Feature selection methods are a branch of data mining approaches to select informative subset of PCS as biomarkers to distinguish a disease from healthy control cohort with high classification accuracy. In this paper, we firstly use three feature selection methods combined with decision tree classification algorithm to select several biomarkers from NEI network. The results show that SVM based decision tree is best fit to select biomarkers for both diseases. Furthermore, we use the theory from Traditional Chinese Medicine (TCM) to divide HL and UA patients into two subgroups. Based on this, we propose a novel feature selection method to distinguish the two subgroups. We combine variance analysis with classification method to select three to four biomarkers for two subgroups in the context of HL and UA respectively, which means that NEI specifications behave differently between two subgroups. According to basic theory of TCM, variant subgroups defined by TCM need to be treated differently. It means that patients with the same disease may be treated in a personalized way. The research efforts in the paper not only to provide a better avenue to understand the nature of diseases, but also to pave a basis to treat two diseases in a personalized way. © 2010 World Scientific Publishing Company. Source

Discover hidden collaborations