Research On Gene Expression Spectrum Recognition Algorithm

With the development of biological big data, the development of microarray technology has made one experiment quantify tens of thousands of gene expression data come true. Biologists have begun to collect large sums of gene expression samples. An urgent problem in the use of microarray data to develop research methods based on the characteristics of gene expression in the samples. The large quantity of gene expression data makes the classification more challenging. The solution is to develop an efficient algorithm to analysis the expression of effective multi-lass data sets.In order to solve the problem of small sample number of large gene data, avoiding the curse of dimensionality, this paper puts forward the method of GESearch, introduces its theory, using four models for gene selection, and make it into a software package for convenience of the use of biologists. The identification and screening of gene expression is visual, people can click to select the desired type of expression spectrum. It is also timeliness and interactivity. Prior knowledge can be brought into the screening of gene expression profiling, which can be better studied on.And the experimental results demonstrate that the expression recognition has advantage in gene recognition, which can classify and recognize the gene better. It also proposes a new subspace learning framework, recursive soft edge(RSM) subspace learning. It is introduced to the gene expression pattern recognition for the dimensionality projection and feature extraction, and prove the convergence. RSM algorithm can make the traditional concept possess soft edges. The use of recursive program is designed to find more projection vector. This algorithm has good effect and application of spectral data in gene expression, and has high accuracy. Experimental results demonstrate that the algorithm has superiority in the identification on gene expression spectrum.