Tumor gene identification method based on gene expression profile

Abstract

The invention provides a tumor gene identification method based on a gene expression profile, which is characterized in that gene expression matrixes of gene samples are studied by adopting a locality preserving projection method and a kernel-function nearest neighbor classification method under the assistance of a computer, the gene expression matrixes of the gene samples are projected into a low-dimensional embedding space so as to reveal a low-dimensional manifold structure hidden in high-dimensional gene expression profile data, and low-dimensional characteristic matrixes are classified by adopting the kernel-function nearest neighbor classification method so that the unrevealed characteristics in the low-dimensional characteristic matrixes are revealed, further the tumor genes in the gene samples are distinguished, and the tumor genes are identified. The method provided by the invention has high rate of identification and good reference value to the clinical diagnosis of the tumor genes and can be applied to establishing a tumor gene identification system.

Description

technical field [0001] The invention relates to the field of computer data processing and gene technology, in particular to a method for identifying tumor genes based on gene expression profiles. Background technique [0002] Cancer has become one of the major diseases that threaten human life. Early detection and diagnosis of cancer is the key to cancer treatment. Tumor gene analysis at the level of gene expression is an important means of diagnosing cancer in the future, while the judgment and identification of tumor genes at the level of gene expression is the premise and basis of tumor gene analysis, which is helpful for the early detection and accurate detection of cancer. judge. [0003] When DNA gene fragments of biological cells are used as gene samples for gene expression analysis, the gene expression profile of the gene sample is usually obtained by gene chip technology, and the gene expression profile is analyzed and studied. However, because the gene sequence f...

AI Technical Summary

Problems solved by technology

However, when used to classify gene expression profiles, both the PCA method and the LDA method have obvious deficiencies, which are mainly manifested in: First, because the overall sample dispersion includes both the intra-class dispersion and the inter-class dispersion of the sample degree, making the PCA method for the purpose of optimal reconstruction not suitable for classification problems; second, although LDA can effectively extract the discriminative information between classes, it is necessary to ensure that the intra-class scatter matrix is ​​reversible during the calculation process, while gene expression The dimensionality of the spectral data is very high, and its intra-class scatter matrix is ​​often singular; third, the PCA method and the LDA method are both obtained under the assumption that the samples obey a multivariate normal distribution. Studies have shown that gene expression The spectrum does not necessarily obey the normal distribution, but is likely to be located on a low-dimensional nonlinear manifold. In this case, the PCA method and the LDA method will probably fail

Images