LightGBM-based circRNA identification method

Abstract

For overcoming defects in prior art, the invention aims to perform identification on the circRNA by means of a lightGBM method. The method comprises the following main steps of (1), inputting the circRNA of a big data sample in a (.bed) file form, wherein the circRNA comprises a chromosome number, a sequence starting site and a positive-and-negative chain mark; (2), mapping the circRNA(.bed) file to a human whole genome (hg19 version) according to the related information of starting site, and obtaining a circRNA sequence information (.fasta) file; (3), presenting a characteristic fusion algorithm for extracting the related characteristic of a circRNA forming process; (4), according to a lightGBM algorithm, respectively performing sample data sampling and characteristic sampling by means of a GOSS and EFB, mapping the continuous characteristics into discrete buckets by means of a Histogram-based algorithm, forming multiple bins, and establishing a histogram by means of the multiple bins, namely discretizing a continuous variable; and (5), adjusting parameters such as a maximal depth max_depth of the tree, a possible minimal recording number min_data_in_leaf of the leaf, and a data proportion bagging_fraction in each iteration, thereby obtaining a model optimal parameter.

Description

technical field [0001] The invention relates to the technical field of bioinformatics, in particular to the field of circRNA identification methods. Background technique [0002] circRNAs have multiple functions in biology, such as being rich in miRNA binding sites and acting as sponges in cells; regulating protein activity by binding to proteins; some circRNAs can even be translated into proteins. Therefore, it has also become a relatively important potential biomarker in recent years. The most important task before discovering the function of circRNA in the experiment is to identify circRNA. However, the time-consuming and high equipment cost of the experimental method with high reliability is not conducive to Identifying circRNAs in large quantities. At the same time, traditional identification tools have problems such as high false positive rate and low overlap rate between multiple methods. It will further negatively affect the functional identification of circRNAs. ...

AI Technical Summary

Problems solved by technology

Therefore, it has also become a relatively important potential biomarker in recent years. The most important task before discovering the function of circRNA in the experiment is to identify circRNA, and the experimental method with a high reliability rate is too time-consuming and expensive, which is not conducive to Identifying circRNAs in large batches

At the same time, traditional identification tools have problems such as high false positive rate and low overlap rate between multiple methods.

Will further negatively affect the functional identification of circRNAs

Images