Abstract:
|
With the advent of precision medicine where the patient treatments are selected based on the genetic understanding of their disease, it is important to know the relationship between genes beforehand. Gene pathways and gene regulatory networks are being constructed to describe the causal relationship between genes, based on biological experiments. However, the pathways and networks for all the known genes are yet to be done and their progression is slow. To address the lack of pathways and networks, gene clustering has been used to group correlated genes together, based on their expression level. While gene expression can be measured at several levels (DNA, RNA, proteins, etc.), most existing methods cluster genes based on one type of omics data (either DNA, or RNA or proteins expression), largely ignoring useful information that can be seen in the other type. We propose a gene clustering method based on multi-omics data using canonical correlation and weighted correlation network analysis. Simulations are run to evaluate the performance of the proposed method and the TCGA breast cancer dataset is used for an illustration.
|