The Algorithm Design Of Risk Factors Identification In Network Structure

The fatigue failure of mechanical parts has a deadly influence to the normal operation of machine. The research of fatigue influence factors is significant to reveal the fatigue mechanism of mechanical parts. The interaction and relationship of fatigue influence factors is complex. So incorporating their synergy and quantitative analysis of influence factors in the fatigue research is a meaningful work. Considering the similarities of fatigue strength’s influence factors and bioinformatics analysis of muti-factors in tumor. The research start from the system analysis to research the identification of differential activity of pathways, we proposed a new effective identify method. Finally, this system analysis method is applied the analysis of mechanical fatigue strength’s influence factors. The main work can be summarized as follows:(1) In this paper, we model gene links for identifying pathways that are associated with survival time of cancer patients. Traditionally, Cox proportional hazard model (CPHM), originally proposed by D. R. Cox in1972, is commonly used to identify clinical features relevant to survival time in cancer research. CPHM can analyze censored data and is independent on the distribution of survive time. We propose a new method, Link-Cox, which combines CPHM with gene links to exploit pathway topological information for survival time-related pathway analysis. Briefly speaking, we represent the activity of a gene link as the production of expression levels of the two genes linked by it and estimate the association of the gene link to survival time based on CPHM. Following this, a new statistic for measuring differential expression of pathways is devised by summarizing significant gene links in pathways. The statistic allows for exploring the co-expression or functional relation of two neighbor genes in a pathway network so that it is biologically reasonable to be used to identify survival-related pathways.(2) To evaluate the proposed method, we first synthesized simulation data and compared it with previous methods based on the simulation data. The experiment results indicates that the performance of accuracy and true positive rate of our methods is better than other methods. Secondly, considering that lung cancer is one of the most malignant tumors world-wide, the proposed method was also verified on real-world lung adenocarcinoma (LUAD) gene expression data sets downloaded from GEO database for identifying pathway signatures for survival time of LUAD patients. In the real-world data evaluation, pathways from KEGG database were collected and used. Extensive comparisons with previous methods show the effectiveness and efficiency of the proposed method for survival pathway hunting.(3) As the proposed method Link-Cox has a better performance in the simulation data set and real data set. We applied the Link-Cox model in the evaluation of the factors which influence the fatigue life of mechanical parts. the fatigue life influence factor includes size, manufacture and notch. The analysis result is consistent with the fatigue experiment. The influence factor, notch, is most significant with the association of fatigue life. So the analysis result demonstrated the risk factors identification method have big application prospect in the analysis of fatigue life influence factors. The risk factors identification method can evaluate the degree of factors impotence quantificationally. It is necessary and very importance for the further research of fatigue life influence factors.