MegaGTA: a sensitive and accurate metagenomic gene-targeted assembler using iterative de Bruijn graphs

Area of Science:

• Metagenomics
• Bioinformatics
• Computational Biology

Background:

• Xander, a gene-targeted metagenomics assembler, utilizes Hidden Markov Models (HMMs) and de Bruijn graphs but has limitations in speed, k-mer size flexibility, and handling of false positives from Bloom filters.
• Xander's reliance on a single k-mer size compromises either sensitivity or accuracy, and its use of Bloom filters introduces potential assembly quality issues due to false positives.

Purpose of the Study:

• To develop a novel gene-targeted assembler, MegaGTA, that addresses the limitations of existing methods like Xander.
• To enhance the sensitivity, accuracy, and computational speed of metagenomic gene assembly.

Main Methods:

• MegaGTA employs iterative de Bruijn graphs to leverage multiple k-mer sizes, optimizing both sensitivity and accuracy.
• Utilizes succinct de Bruijn graphs (SdBG) for an exact representation of the de Bruijn graph, reducing memory footprint and enabling faster computation through parallel algorithms.
• Incorporates k-mer multiplicity for improved Hidden Markov Model (HMM) construction, avoiding false-positive contigs.

Main Results:

• MegaGTA demonstrated superior sensitivity and accuracy compared to Xander on mock metagenomic datasets.
• Analysis of a large rhizosphere soil metagenomic sample showed MegaGTA produced 9.7–19.3% more contigs and assigned 10–25% more gene references than Xander.
• MegaGTA exhibited a 2- to 10-fold increase in speed over Xander, depending on the number of k-mers used.

Conclusions:

• MegaGTA significantly improves upon Xander's gene assembly algorithms, delivering enhanced sensitivity, accuracy, and speed.
• The assembler is capable of efficiently assembling gene sequences from ultra-large metagenomic datasets.
• MegaGTA's source code is publicly available for further research and application.