Gene prediction in metagenomic fragments based on the SVM algorithm

Abstract

Background: Metagenomic sequencing is becoming a powerful technology for exploring micro-ogranisms from various environments, such as human body, without isolation and cultivation. Accurately identifying genes from metagenomic fragments is one of the most fundamental issues. Results: In this article, we present a novel gene prediction method named MetaGUN for metagenomic fragments based on a machine learning approach of SVM. It implements in a three-stage strategy to predict genes. Firstly, it classifies input fragments into phylogenetic groups by a k-mer based sequence binning method. Then, proteincoding sequences are identified for each group independently with SVM classifiers that integrate entropy density profiles (EDP) of codon usage, translation initiation site (TIS) scores and open reading frame (ORF) length as input patterns. Finally, the TISs are adjusted by employing a modified version of MetaTISA. To identify protein-coding sequences, MetaGun builds the universal module and the novel module. The former is based on a set of representative species, while the latter is designed to find potential functionary DNA sequences with conserved domains.

A New Feature Selection Methodology for K-mers Representation of DNA Sequences

“…K-mer representation have shown its effectiveness in several in-silico analysis applied to different genomics and epigenomics studies. In particular they have been used to characterize nucleosome positioning [4], to find enhancer functional regions [5], to characterize epigenetic variability [6], in sequence alignment and transcriptome assembly [7] and in gene prediction [8]. The interested reader can find the basic ideas of k-mer based methods to different biological problems in the following review [17].…”

A Comprehensive Review of Emerging Computational Methods for Gene Identification

“…Different from linear discriminant functions, non-liner kernels have complex discriminant functions for complicated data examples. Usually, classical non-linear kernels designed for particular applications, including polynomial kernels [76], Gaussian kernels [79,80], spectrum kernels [81], weighted degree (WD) kernels [74], WD kernels with shifts [82], string kernels [83,84], Oligo kernels [85], convolutional kernels [86], and so forth, can be used for modeling more complex decision boundaries in predicting various signal sensors [72,74,87].…”

“…3) Metagenomic Sequences: Metagenomic sequencing has emerged as a powerful tool for exploring environmental organisms without isolation and cultivation [80]. The shotgun sequencing of microbial communities generates numerous metagenomic short transcriptional sequences that are heterogeneous and mixed together.…”

“…Also, k-mer-based sequence binning methods and sequence property-based methods are often seen as the input for training models. For example, a newly developed SVM-based algorithm [80] can be implemented in a three-stage strategy to predict genes. First, short reads are classified into phylogenetic groups by a k-mer-based sequence binning method.…”

N‐terminomics and proteogenomics, getting off to a good start

“…In addition to all of the difficulties faced by gene prediction in single-strain genomes, gene prediction in metagenomics is further complicated by the read length, which is often short, meaning that the upstream regions of coding sequences may be missing, and the fact that the individual genomes contributing to the metagenome are often unknown [24]. The accuracy of metagenomic gene predictions is difficult to gauge given the lack of references; experimentally verified translation initiation sites are few and far between in metagenomic samples [24]. Similar to translation initiation site prediction, many algorithms for signal peptide cleavage site prediction rely on machine learning to recognize the surrounding context [25].…”