Identification Of Cancer Biomarkers Based On The Perspective Of Complex Network Analysis Method And Survival Prediction Analysis Of Cancer

Tumor markers refer to substances that can be detected in the blood,body fluids and tissues that are related to tumorigenesis and progression.The presence or quantification of tumor markers can indicate the nature of tumors,so as to be used to understand the occurrence and differentiation of tumors.It plays an important role in diagnosis of cancer,relapse,prognosis and other aspects.As tumor markers,Micro RNA(miRNA)and protein play important roles in the development and progression of the tumor,few researches have been conducted to determine their effects on progress of tumor from network perspective.This paper presents a complex network analysis method to study the miRNA and protei n associated with breast cancer patients.The miRNA expression level network,structure network and sequence network are constructed according to the Euclidean distance between miRNA expression levels,the difference between secondary structures and the s equence alignment of miRNA.The relationship between miRNA sequence,secondary structure and expression level is studied based on the complex network statistics such as average edges overlap and degree distribution.Random forest model is employed for filtering and wrapping miRNAs and proteins from breast cancer patients' normal and tumor tissues in order to select a subset of smallest number of miRNA or protein and to maintain the highest classification accuracy.Maximal Information Coefficient(MIC)of an y pair of miRNA is then used as distance to determine their connection in network.Comparing the miRNAs network from normal tissue and tumor tissue,eight miRNA in two networks are identified according to the significant differences in node betweeness.Namely,hsa-mir-101-2,hsa-mir-10 b,hsa-mir-130 b,hsa-mir-193 a,hsa-mir-204,hsa-mir-28,hsa-mir-365-2,hsa-mir-375,in particular eight miRNA have been highly related with breast cancer.We also applied this proposed method to the case of prostate cancer.Accordingly,seven of eight selected miRNA are related with prostate cancer.Minimal net clustering group miRNA with similar sequence into a same class,miRNA in the same cluster,half of them have similar functions on cancer,such as promote proliferation,invasion of cancer cells.It is supposed that unverified miRNA which has dramatic difference of node betweenness in two networks might also be associated with the progression of breast cancer.We call the above method of selecting tumor biomarkers as compl ex network analysis method.This method shows its feasible and for discovering tumor biomarkers associated with cancer occurrence and progress.The regression relationship between miRNA,protein,stage of cancer and the survival of patients with breast cancer was established by stepwise multiple regression method.Survival was predicted by fewer independent variables.Selecting variables that have a greater impact on survival,hsa-mir-548,hsa-mir-943,CD20-RC,e IF4G-RC,GSK3_p S9-RV and cancer staging are found to have a strong correlation with the survival of cancer patients.