MG-RAST

MG-RAST

Original author(s)	Argonne National Laboratory, University of Chicago, San Diego State University
Developer(s)	F. Meyer, D. Paarmann, M. D'Souza, R. Olson, E.M. Glass, M. Kubal, T. Paczian, A. Rodriguez, R. Stevens, A. Wilke, J. Wilkening, R.A. Edwards
Initial release	2008 (2008)
Stable release	4.0 / 15 November 2016 (2016-11-15)
Type	Bioinformatics
Website	http://metagenomics.anl.gov/

MG-RAST, an open-source web application server, facilitates automatic phylogenetic and functional analysis of metagenomes. It stands as one of the largest repositories for metagenomic data, employing the acronym for Metagenomic Rapid Annotations using Subsystems Technology (MG-RAST). This platform utilizes a pipeline that automatically assigns functions to metagenomic sequences, conducting sequence comparisons at both nucleotide and amino acid levels. Users benefit from phylogenetic and functional insights into the analyzed metagenomes, along with tools for comparing different datasets. MG-RAST also offers a RESTful API for programmatic access.

Argonne National Laboratory from the University of Chicago created and maintains this server. As of December 29, 2016, MG-RAST had analyzed a substantial 60 terabase-pairs of data from over 150,000 datasets. Notably, more than 23,000 of these datasets are publicly available. Computational resources are currently sourced from the DOE Magellan cloud at Argonne National Laboratory, Amazon EC2 Web services, and various traditional clusters.

Background

MG-RAST was developed to serve as a free, public resource dedicated to the analysis and storage of metagenome sequence data. It addresses a key bottleneck in metagenome analysis by eliminating the dependence on high-performance computing for annotating data.

The significance of MG-RAST becomes evident in metagenomic and metatranscriptomic studies, where the processing of large datasets often requires computationally intensive analyses. With the substantial reduction in sequencing costs in recent years, scientists can generate vast amounts of data. However, the limiting factor has shifted to computing costs. For example, a recent University of Maryland study estimated a cost exceeding $5 million per terabase using their CLOVR metagenome analysis pipeline. As sequence datasets' size and number continue to grow, the associated analysis costs are expected to rise.

Beyond analysis, MG-RAST functions as a repository tool for metagenomic data. Metadata collection and interpretation are crucial for genomic and metagenomic studies. MG-RAST addresses challenges related to the exchange, curation, and distribution of this information. The system has embraced minimal checklist standards and biome-specific environmental packages established by the Genomics Standards Consortium. Furthermore, MG-RAST provides a user-friendly uploader for capturing metadata at the time of data submission.

Pipeline for metagenomic data analysis

The MG-RAST application provides a comprehensive suite of services, including automated quality control, annotation, comparative analysis, and archiving for metagenomic and amplicon sequences. It utilizes a combination of various bioinformatics tools to achieve these functionalities. Originally designed for metagenomic data analysis, MG-RAST also extends support to amplicon sequences (16S, 18S, and ITS) and metatranscriptome (RNA-seq) sequences processing. However, it's important to note that MG-RAST currently lacks the capability to predict coding regions from eukaryotes, limiting its utility for eukaryotic metagenome analysis.

The MG-RAST pipeline can be segmented into five distinct stages:

Data hygiene

The MG-RAST pipeline incorporates a series of steps for quality control and artifacts removal, ensuring robust processing of metagenomic and metatranscriptome datasets. The initial stage involves trimming low-quality regions using SolexaQA and eliminating reads with inappropriate lengths. In the case of metagenome and metatranscriptome datasets, a dereplication step is introduced to enhance data processing efficiency.

The subsequent step employs DRISEE (Duplicate Read Inferred Sequencing Error Estimation) to evaluate sample sequencing errors by measuring Artificial Duplicate Reads (ADRs). This assessment contributes to enhancing the accuracy of downstream analyses.

Finally, the pipeline offers the option to screen reads using the Bowtie aligner. It identifies and removes reads that exhibit matches close to the genomes of model organisms, including fly, mouse, cow, and human. This step aids in refining the dataset by filtering out reads associated with potential contaminants or unintended sequences.

Feature extraction

In the gene identification process, MG-RAST employs a machine learning approach known as FragGeneScan. This method is utilized to identify gene sequences within the metagenomic or metatranscriptomic data.

For the identification of ribosomal RNA sequences, MG-RAST initiates a BLAT search against a reduced version of the SILVA database. This step allows the system to pinpoint and categorize ribosomal RNA sequences within the dataset, contributing to a more detailed understanding of the biological composition of the analyzed metagenomes or metatranscriptomes.

Feature annotation

To identify the putative functions and annotations of the genes, MG-RAST follows a multi-step process. Initially, it builds clusters of proteins at a 90% identity level using the UCLUST implementation in QIIME. The longest sequence within each cluster is then selected for further analysis.

For the similarity analysis, MG-RAST employs sBLAT, a parallelized version of the BLAT algorithm using OpenMP. The search is conducted against a protein database derived from the M5nr, which integrates nonredundant sequences from various databases such as GenBank, SEED, IMG, UniProt, KEGG, and eggNOGs.

In the case of reads associated with rRNA sequences, a clustering step is performed at a 97% identity level. The longest sequence from each cluster is chosen as the representative and is used for a BLAT search against the M5rna database. This database integrates sequences from SILVA, Greengenes, and RDP, providing a comprehensive reference for the analysis of ribosomal RNA sequences.

Profile generation

The data feeds several key products, primarily abundance profiles. These profiles summarize and reorganize the information found in the similarity files in a more easily digestible format.

Data loading

Finally, the obtained abundance profiles are loaded into the respective databases.

Detailed steps of the MG-RAST pipeline

MG-RAST Pipeline	Description
qc_stats	Generate quality control statistics
preprocess	Preprocessing, to trim low-quality regions from FASTQ data
dereplication	Dereplication for shotgun metagenome data by using k-mer approach
screen	Removing reads that are near-exact matches to the genomes of model organisms (fly, mouse, cow and human)
rna detection	BLAT search against a reduced RNA database, to identifies ribosomal RNA
rna clustering	rRNA-similar reads are then clustered at 97% identity
rna sims blat	BLAT similarity search for the longest cluster representative against the M5rna database
genecalling	A machine learning approach, FragGeneScan, to predict coding regions in DNA sequences
aa filtering	Filter proteins
aa clustering	Cluster proteins at 90% identity level using uclust
aa sims blat	BLAT similarity analysis to identify protein
aa sims annotation	Sequence similarity against protein database from the M5nr
rna sims annotation	Sequence similarity against RNA database from the M5rna
index sim seq	Index sequence similarity to data sources
md5 annotation summary	Generate summary report md5 annotation, function annotation, organism annotation, LCAa annotation, ontology annotation and source annotation
function annotation summary	Generate summary report md5 annotation, function annotation, organism annotation, LCAa annotation, ontology annotation and source annotation
organism annotation summary	Generate summary report md5 annotation, function annotation, organism annotation, LCAa annotation, ontology annotation and source annotation
lca annotation summary	Generate summary report md5 annotation, function annotation, organism annotation, LCAa annotation, ontology annotation and source annotation
ontology annotation summary	Generate summary report md5 annotation, function annotation, organism annotation, LCAa annotation, ontology annotation and source annotation
source annotation summary	Generate summary report md5 annotation, function annotation, organism annotation, LCAa annotation, ontology annotation and source annotation
md5 summary load	Load summary report to the project
function summary load	Load summary report to the project
organism summary load	Load summary report to the project
lca summary load	Load summary report to the project
ontology summary load	Load summary report to the project
done stage
notify job completion	Send notification to user via email

MG-RAST utilities

G-RAST isn't just a powerhouse for metagenome analysis, it's also a treasure trove for data exploration. Dive into a diverse toolbox for visualizing and comparing metagenome profiles across various datasets. Filter based on specifics like composition, quality, functionality, or sample type to tailor your search. Delve deeper with statistical inferences and ecological analyses – all within the user-friendly web interface.

See also

• Metagenomics

External links

• MG-RAST Web Server
• API
• MG-RAST manual
• M5NR