A Functional Module Detection Method By Combining Cancer Multi-omics Data And PPI Network

Cancer is a complex disease that seriously threatens human health.Its pathogenic mechanism is complex and difficult to cure.Studies have found that the occurrence and development of cancer is the result of a combination of various factors.With the development of high-throughput sequencing technology,multi-omics data is becoming more and more abundant.How to use massive multi-omics data to obtain disease-related information is currently a hot research field.Among them,DNA methylation and gene expression are important factors leading to cancer.As an important epigenetic mechanism,abnormal DNA methylation often regulates gene expression levels,which may lead to cancer.As a product of gene expression,protein cannot play its biological role alone.It needs to interact with other molecules to complete some complex physiological functions.The Protein-Protein Interaction(PPI)network is a biomolecular network that plays an important role in biological activities.Nowadays,although the sequence order of most proteins is known,their molecular functions are still not well explained.Research on functional modules of PPI networks helps to understand biological mechanisms and with the continuous advancement of computer technology,more and more researchers use machine learning and data mining related algorithms to process PPI network data,in order to find out the cancer-related pathogenic gene function modules and further understand its pathogenic mechanism.In this paper,DNA methylation and gene expression association patterns of various cancers were studied at the genome-wide level,then genes with differential expression and differential methylation were mapped to PPI networks.Next we get cancer-related pathogenic gene function modules by using an overlapping functional module discovery method.The main innovative achievements of the paper are as follows:1.In view of the systematic errors in PPI network data and the incomplete data due to biological experiment problems,this study combines PPI network data with multi-omics data.It makes a variety of information complementary,which is conducive to mining deep patterns.Gene expression data and DNA methylation data were first processed,and a π-value based differential analysis method was used to obtain genes with abnormal methylation levels and gene expression abnormalities,and further analyzed by fusion with PPI network.The method makes the constructed network have some priori information,which effectively compensates for the incomplete features of the PPI network.2.In this paper,a new DKNMF-based overlapping module discovery method is proposed.Based on the construction network,the diffusion kernel function is used to calculate the feature matrix.The matrix not only indicates the association between connected nodes,but also indicates the association between indirectly connected nodes.Then,the non-negative matrix decomposition method is used to perform multiple iteration calculations,and the optimal division number is obtained according to the modularity value,so the membership degree matrix of the gene is further obtained..At the same time,the dense subnet modules are selected from the overlap between modules and the module density.Compared with other functional module discovery methods,the experimental results show that the proposed method has better performance.3.This paper conducted experimental studies on multiple cancer data sets.We analyzed DNA methylation data and gene expression data of each cancer,and combined with PPI network to find out cancer-related functional modules and disease-causing genes.The experimental results show that the obtained functional modules are closely connected internally,the connections between the modules are sparse,also functional modules with overlapping structures can be obtained.Then,the enrichment analysis results of functional modules from a variety of cancers have good biological interpretations,and the enriched signaling pathways are directly or indirectly related to cancer.The above results show that the proposed method does not depend on the data set,so it can be applied to other fields of research and has good scalability.In conclusion,the proposed DKNMF-based overlapping functional module detection method can obtain cancer-related gene clusters from the whole genome level,which is helpful for complex disease research and provides a theoretical basis for the diagnosis,treatment and prognosis of cancer.