Research On Validation Methods Of Cancer Biomarkers Based On Gene Expression Profile And Knowledge Base Development

At present the fields of life sciences,clinical medicine,medical and bioinformatics are all focused on the research of biomarkers.Biomarkers play a key role in clinical diagnosis and prognosis analysis,and the development of new drugs.From the perspective of OMICs research,classification and statistical inference techniques are usually used to mine the differentially expressed genes and discovery the potential risk factors.The eyes of Artificial Intelligence are always on constructing the expert system or knowledge base,by mining the massive documents and forming the structural knowledge.Based on the model-driven methodology,here the paper designed and developed an gene expression profile validation validation and knowledge base system.The main work as following:(1)The survival analysis techniques of gene biomarkers were studied.The effect of gene expression on the survival time of sample observations was analyzed based on gene expression profile and clinical annotations with censored data.Using the log-rank test,the optimal threshold of gene expression for risk grouping was determined by the minimum P-value criteria.Since multiple tests are involved,the minimum P-value was corrected using different methods.Cox proportional hazards models were used to verify whether a gene has significant effect on survival.(2)Analyzed how to use RDF to describe the information of gene biomarkers,and how to map RDF entities and relations to a property graph model.Determined the target information of gene biomarkers that will be stored in the knowledge base.The template of the Web page from the source server was analyzed,an information extraction crawler and extraction tool was designed and implemented to extract the gene biomarkers information from OMIM database and Gene Cards database.(3)Designed and developed a gene biomarkers validation and knowledge base system.A batch of gene expression profiles and corresponding clinic annotation data ware preprocessed and integrated to the system.The system also provide friendly visualization of the validation result and the gene related information.