Metagenome sequence deep clustering method based on reference species tag constraint

Abstract

The invention provides a metagenome sequence deep clustering method based on reference species tag constraints. A deep learning pre-training model based on the reference species tag constraints is designed. According to the method, a pre-training database of known species based on different communities is established, each 4mer feature vector is divided into three conditions of the same species, different species of the same genus and different species of the different genus when the pre-training database is established, and relationships among 4mer features of sequences among samples under the three conditions are researched respectively; a label constraint error function of a pre-training model is established, pre-training is performed by using a database of known labels of the community, and different pre-training models are constructed for different microbial communities; when a user uses the method, the user only needs to load the pre-training model of the required community for different communities, and the clustering result can be obtained only by waiting for iteration of the fine tuning step for several times when the model is reloaded. Finally, the clustering method can show very excellent clustering performance.

Description

technical field [0001] The invention relates to the field of bioinformatics analysis, in particular to a method for deep clustering of metagenomic sequences based on reference species label constraints. Background technique [0002] Microorganisms are the largest, most numerous, and most widely distributed group of organisms on earth. People's research on microorganisms is mainly based on pure culture, but it was later found that more than 99% of microorganisms are not cultivable. In order to study microorganisms that cannot be cultivated, a new concept - metagenomics came into being. Metagenomics uses next-generation sequencing technology to obtain most of the genetic material in the environment without laboratory cultivation. Different from traditional sequencing methods, the raw data obtained by metagenomic sequencing are a large number of short DNA fragments derived from a variety of microorganisms. According to the overlapping relationship between DNA fragments, rese...

AI Technical Summary

Problems solved by technology

[0005] In view of the deficiencies in the prior art above, the purpose of the present invention is to provide a method for deep clustering of metagenomic sequences constrained by reference species labels, aiming to solve the problem of the same genus when performing clustering of metagenomic DNA sequences in the prior art. The problem of inaccurate clustering caused by the high similarity of adjacent species

Images