Tumor key gene identification method based on prior information and parallel binary particle swarm optimization

Abstract

The invention discloses a tumor key gene identification method based on prior information and a parallel binary particle swarm optimization. The tumor key gene identification method comprises the steps of performing preprocessing on tumor gene expression profile data, on a training set, determining an optimal gene cluster number K through a user-defined criterion function by an improved Elbow method; preferably selecting K optimal cluster centers by the particle swarm optimization (PSO), and clustering the tumor genes into K classes on the training set by a K-mean value method; on the training set, obtaining gene to class sensitivity (GCS) information and gene regulation (GR) information separately; and taking the obtained K gene clusters as a searching spacing, by combining with the obtained two kinds of prior information, identifying the key tumor genes by adopting the parallel binary particle swarm optimization (BPSO). Compared with the existing tumor key gene identification method, the probability of losing key information genes related to classes of tumors is lowered by consideration of the two kinds of prior information in the tumor key gene identification method disclosed by the invention, so that subsequent tumor identification can be improved.

Description

technical field [0001] The invention belongs to the application field of computer analysis technology of tumor gene expression spectrum data, and specifically relates to a tumor key gene identification method based on prior information and parallel binary particle swarm algorithm. Background technique [0002] In life science research, while DNA microarray technology brings unprecedented opportunities for biological and medical research, the complex gene expression profile data it generates poses great challenges to existing data analysis and processing methods. First, gene expression profile data has very high dimensions (genes), and there are very complex relationships between these gene dimensions. Second, gene expression profile data have fewer samples, which contradicts the huge number of genes that constitute an imbalance. Third, gene expression profile data inherently have data analysis difficulties such as high noise and high variation. Fourth, a large amount of us...

AI Technical Summary

Problems solved by technology

Second, gene expression profile data has fewer samples, which contradicts the huge number of genes that constitute an imbalance

Third, gene expression profile data inherently have data analysis

Images