Unsupervised clustering method and system for metagenome contigs

Abstract

The invention discloses an unsupervised clustering method for metagenome contigs, which comprises the following steps of: before clustering, putting reads of each sample together to construct a gene library, assembling the reads into contigs by using an assembling tool, and performing feature vectorization on each contigs according to the frequency of quadruple oligonucleotides and co-abundance, then carrying out clustering according to a pre-trained probability model and a recursion strategy, introducing CheckM to carry out clustering result quality detection and simplify the complexity of asample and taking the clustering result as an algorithm termination condition, and taking a result obtained by Marker gene analysis as strain number initialization and clustering center sequence initialization in algorithm initialization.

Description

technical field [0001] The invention relates to the technical field of biological information analysis, in particular to an unsupervised clustering method and system for metagenomic contigs. Background technique [0002] Before the emergence of metagenomic technology, people's research on microorganisms was mainly through the pure cultivation of a single microbial species. However, in the natural environment, most microorganisms are difficult or impossible to be purely cultured on the culture medium. With the development of second-generation sequencing technology, metagenomic technology came into being. It can directly obtain the genetic material of all microorganisms in the sample from the natural environment, without the need for pure culture on the medium like traditional methods. This provides new research ideas for scientists to study the community structure of microorganisms, the interaction between microorganisms, and the relationship between microorganisms, the env...

AI Technical Summary

Problems solved by technology

The clustering methods included in DAS tool and Metawrap are applied to the same data set at the same time. Some methods do not perform well in complex data sets (such as CONCOCT, MyCC). At the same time, the ability of most methods to reconstruct genes on complex data sets needs to be improved. Therefore, although the integration of the clustering results of these methods can improve the number and quality of reconstructed genes, the effect is limited.

[0004] The main problems of the existing unsupervised clustering methods are: 1. The ability to reconstruct genes in complex environments needs to be improved

2. The selection of the number of strains in the clustering process is very different from the real situation. The number of strains is a key parameter of the unsupervised clustering method, which greatly affects the performance of the algorithm

3. It is difficult to distinguish the same species but different strains in the sample. There are two reasons. One is to use the assembly tool to assemble the reads. Due to the high similarity between the same species but different strains, chimeras are prone to occur during the assembly process , the second reason is that there are a large number of strains in a complex environment, and it is difficult to effectively distinguish

Images