Marker screening method based on single variable and pair variables

Abstract

The invention belongs to the technical field of biological data analysis, and relates to a marker screening method based on a single variable and pair variables. As biological data has the advantagesof being high in dimensionality and small in sample size, variables are evaluated and selected through the simple and accurate decision rule, and classification and forecasting are important tasks ofbiological data analysis. For comprehensively evaluating the variables, a best split point of the single variable is calculated through information gain. A new variable is constructed through the bestsplit point, and the pair variables are established through the new variable and an original variable corresponding to the new variable. Meanwhile, variables of primal space generate variable pairs through combination two by two. All the pair variables are scored according to two scoring criterions, and are sorted in descending order according to scores, k pairs of variables which have the highest score and are not overlapped are selected to establish a fusion classifier. According to the marker screening method based on the single variable and the pair variables, the new variable is constructed through the single variable, the classification performance of the single variable and the classification performance of the pair variables can be evaluated through the same rules, and the and effective data processing method is provided.

Description

technical field [0001] The invention belongs to the technical field of biological data analysis, relates to a marker screening method based on univariate and pair variables, and is a feature selection and classification method for simultaneously measuring univariate and pair variables. Background technique [0002] Biological data usually has the characteristics of high dimensionality and small sample size, so using simple and accurate decision rules to evaluate and select variables to classify and predict is an important task of biological data analysis, which has great significance for the research of disease diagnosis, drug efficacy, prognosis, etc. important meaning. [0003] Single-molecule markers are often used as important indicators for clinical diagnosis and prognosis. For example, alpha-fetoprotein (AFP) has been considered the preferred serum tumor marker for the diagnosis of liver cancer. However, molecular individual markers often suffer from high false positi...

AI Technical Summary

Problems solved by technology

However, molecular individual markers often suffer from high false positives

Images