Cancer hypotype biomarker detecting system based on student t distribution

Abstract

The invention relates to a cancer hypotype biomarker detecting system based on student t distribution. The cancer hypotype biomarker detecting system comprises a gene expression profile extracting unit, a network cluster model establishing unit, a cancer hypotype finding unit, an intergenic mutual relation exploring unit and a biomarker detecting unit. The gene expression profile extracting unit is used for extracting an expression profile of a gene sample in a gene chip. When the extracted gene expression profile is input to the network cluster model establishing unit, the network cluster model establishing unit is used for establishing a mixture model and solving the gene expression profile, wherein the mixture model uses the diverse student t distribution to describe each component. The cancer hypotype finding unit is used for finding new cancer hypotypes by utilizing the mixture model and a clustering method. The intergenic mutual relation exploring unit regards each gene as a point inside a network, and is used for exploring the mutual relation among genes in different cancer hypotypes by exploring the network structure. The biomarker detecting unit is used for detecting a biomarker related to the cancer hypotypes according to the mutual relation among the genes in different cancer hypotypes. The cancer hypotype biomarker detecting system describes each component in the mixture model by using the diverse student t distribution, and can have robustness for abnormal points and noise in the clustering and variable selecting process.

Description

technical field [0001] The present invention relates to the fields of bioinformatics, statistical learning and pattern recognition, and more specifically, relates to a cancer subtype biomarker detection system based on Student's t distribution. Background technique [0002] With the increasing amount of genome expression profiling data, microarray methods have become a key technology for identifying cancer-related genes and discovering new cancer subtypes (subclasses). Compared with clinical and pathological risk factors such as patient age, tumor size, and steroid receptor status, the study of genes can provide a deeper understanding of the physiology of cancer and more effectively detect new cancer subtypes. These cancer subtypes may differ in gene or protein expression profiles, gene regulation, or protein signaling networks. Detecting these cancer subclasses from gene expression profiles can be viewed as an unsupervised clustering problem, while finding genes for detect...

AI Technical Summary

Problems solved by technology

Additionally, systems using the diagonal covariance assumption ignore correlations between genes in cancer subtypes

Moreover, the traditional penalty-based model clustering algorithms all use the mean value as the criterion for selecting genes, while ignoring their influence on networks and paths, which is very unfavorable for understanding the collective biological functions of genes

Images