Pathway Functional Network Construction And Analysis Based On Semantic Similarity Of Gene Ontology

The expression of genes and proteins varies from cells in different tissues and different stage of development, and it is still different even in the same cell by different stimulation. Recent years, with the rapid and sustained development of high throughput and large scale sequencing technology, coupled with the increasing improvement of microarray, it makes easier and more convenient for researchers to acquire the expression of human genes or proteins at different times and states. The emergence and development of these techniques have helped researchers accumulate large amounts of data, and it is urgently needed to find out the critical pathways related to these data in order to provide theoretical basis for exploring the pathogenesis, discovering of accurate disease treatment and making clinical individualized medication. Traditional differential expression gene screening and pathway annotation methods can find the disease related risk pathways, but disable to explore and analyze the associations between pathways, and can’t elucidate how they interact with each other and then affect the development of the disease.Biological molecular network is a powerful tool to study the relationship between molecular components. It has been widely used in exploring the mechanism of complex diseases. So we hope to build a functional network which can effectively describe the relationship of pathways in certain disease, and provide help to explore the mechanism of disease.In this paper, we proposed a method to construct the pathway functional network. Based on the network constructed, we can analyze the attributes and sub-structures of the network and find out disease pathogenesis pathways. The specific process of constructing the network includes: firstly, the semantic similarity between different differential expressed genes is calculated using Gene Ontology knowledge structure. Secondly, based on the similarity between differential expressed genes, we calculate the correlation between the pathways, and then construct the pathway network. At last, the critical pathways are identified and key modules are mined in the network. Compared with DAVID and other pathway annotation methods, the results on breast cancer data show that our method can not only find the critical pathway of high pathogenic risk, but also find the relationships between them, which provides help and guidance for gaining insight into the mechanism of disease.