Data correction and classification method and storage medium

Abstract

The invention discloses a data correction and classification method and a storage medium. The data correction and classification method comprises a sequence data acquisition step, a data calibration step, a data reconstruction step, a classification step, a discriminator training step and a target batch label vector speculation step. The correction of the batch effect is realized, and the problem that the distribution heights of two batches of data are not matched is solved. An end-to-end combined deep learning framework is provided to classify sequence data on the basis of data correction, the framework is verified on the data of a flow cytometry and a laser desorption ionization mass spectrum, and especially for the latter, the diagnosis accuracy is greatly improved. Compared with the current most advanced mainstream method, the average value is improved by about 5.57.9%. Experiments prove that the performance of the developed method is obviously superior to that of a traditional method, and the influence of the batch effect is overcome.

Description

technical field [0001] The invention relates to the field of classification of metabolomics data based on artificial intelligence methods, and in particular to a data correction and classification method and a storage medium. Background technique [0002] Metabolomics is an important branch of the five omics analysis, and plays an important role in both clinical application and basic research on metabolic biomarkers. However, related research in this field is often affected by batch effects due to many external factors. [0003] In the past two decades, the bottleneck caused by the batch effect has attracted widespread attention of many scholars in the industry, and many algorithms for eliminating the batch effect have been developed. According to their working principles, these traditional techniques can be divided into two categories: methods based on location scale (LS) and methods based on matrix factorization (MF). The first category includes Empirical Bayesian method...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the invention is to provide a method to solve the problem of different platforms, different reagents of the same sample and different times of data acquisition due to the influence of batch effects when extending machine learning models to actual metabolomics data. Repeated processing of patient serum samples on different target plates leads to inconsistent diagnostic results, which seriously hinders its application to machine learning for classification and technical problems in diagnosis

Images