IPMM: Cancer Subtype Clustering Model Based on Multiomics Data and Pathway and Motif Information

Multiomics compiles data from different genome levels to study the effects of interactions between various omics molecules on disease processes. Integrated analysis of different omics data can more comprehensively evaluate their role in human health and complex diseases. Previous studies have used SNF and SNF-CC for multiomics integration. Although the effect of multiomics integrative algorithm is significantly increased, these methods did not consider the effects of a biologically significant correlation within and between omics. A large body of evidence has shown that cancer occurs due to interactions and synergistic effects of multiple genes. The correlation relationships between genes can be reflected through gene pathway and motif information. In this paper, we define the IPMM(Integration Pathway and Motif information Model), which combines pathway and motif information with multiomics data to study their effects on cancer subtype classification. To facilitate the use of gene association information, we employ the Isomap method for dimensionality reduction analysis of expression data from the genomes in a pathway and motif. Selection of K values in Isomap dimensionality reduction is used to maximize the presentation of the relationship of genes in pathway and motif data with dimensionality reduced to one. SNF and SNF-CC are used for integrative analysis of gene-expression data, methylation data, miRNA data, and pathway and motif data after dimensionality reduction in two cancer datasets. Results show that clustering effects display varying increases in different methods after pathway and motif information are integrated.