ASaiM: a Galaxy-based framework to analyze raw shotgun data from microbiota

Background New generation of sequencing platforms coupled to numerous bioinformatics tools has led to rapid technological progress in metagenomics and metatranscriptomics to investigate complex microorganism communities. Nevertheless, a combination of different bioinformatic tools remains necessary to draw conclusions out of microbiota studies. Modular and user-friendly tools would greatly improve such studies. Findings We therefore developed ASaiM, an Open-Source Galaxy-based framework dedicated to microbiota data analyses. ASaiM provides a curated collection of tools to explore and visualize taxonomic and functional information from raw amplicon, metagenomic or metatranscriptomic sequences. To guide different analyses, several customizable workflows are included. All workflows are supported by tutorials and Galaxy interactive tours to guide the users through the analyses step by step. ASaiM is implemented as Galaxy Docker flavour. It is scalable to many thousand datasets, but also can be used a normal PC. The associated source code is available under Apache 2 license at https://github.com/ASaiM/framework and documentation can be found online (http://asaim.readthedocs.io/) Conclusions Based on the Galaxy framework, ASaiM offers sophisticated analyses to scientists without command-line knowledge. ASaiM provides a powerful framework to easily and quickly explore microbiota data in a reproducible and transparent environment.


Findings Background
The study of microbiota and microbial communities has been facilitated by the evolution of sequencing techniques and the development of metagenomics and metatranscriptomics.
These techniques are giving insight into phylogenetic properties and metabolic components of microbial communities. However, meta'omic data exploitation is not trivial due to the large amount of data, high variability, incompleteness of reference databases, difficulty to find, configure, use and combine the dedicated bioinformatics tools, etc. Hence, to extract useful information, a sequenced microbiota sample has to be processed by sophisticated workflows with numerous successive bioinformatics steps [1] . Each step may require execution of several tools or software programs. For example, to extract taxonomic information with the widely used QIIME [2] or Mothur [3] , at least 10 different tools with at least 4 parameters each are needed. Designed for amplicon data, both QIIME and Mothur can not be directly applied to shotgun metagenomics data. In addition, the tools can be complex to use; they are command-line tools and may require computational resources specially for the metagenomics datasets. In this context, selecting the best tools, configuring them to use the correct parameters and appropriate computational resources and combining them together in an analysis chain is a complex and error-prone process. These issues and the involved complexity are blocking scientist from participating in the analysis of their own data.
Furthermore, bioinformatics tools are often manually executed and/or patched together with custom scripts. These practices raise doubts about a science gold standard: reproducibility [3,4] . Web services and automated pipelines such as MG-RAST [5] and EBI metagenomics [6] offer solutions to the accessibility issue. However, these web services work as a black box and are lacking in transparency, flexibility and even reproducibility as the version and parameters of the tools are not always available. Alternative approaches to improve accessibility, modularity and reproducibility can be found in open-source workflow systems such as Galaxy [6][7][8] . Galaxy is a lightweight environment providing a web-based, intuitive and accessible user interface to command-line tools, while automatically managing computation and transparently managing data provenance and workflow scheduling [6][7][8] .
More than 4,500 tools can be used inside Galaxy environments. The tools can be selected and combined to build Galaxy flavors focusing on specific type of analysis, e.g. the Galaxy RNA workbench [9] .
In this context, we developed ASaiM (Auvergne Sequence analysis of intestinal Microbiota), an Open-Source opinionated Galaxy-based framework. It integrates tools and workflows dedicated to microbiota analyses with an extensive documentation ( http://asaim.readthedocs.org ) and training support.

Goals of ASaiM
ASaiM is developed as a modular, accessible, redistributable, sharable and user-friendly framework for scientists working with microbiota data. This framework is unique in combining curated tools and workflows and providing easy access for scientists.
ASaiM is based on four pillars: 1) easy and stable dissemination via Galaxy, Docker and conda, 2) a comprehensive set of metagenomics related tools, 3) a set of predefined and tested workflows, and 4) extensive documentation and training to help scientists in their analyses.

A framework built on the shoulders of giants
The ASaiM framework is built on existing tools and infrastructures and combine all their forces to build an easily accessible and reproducible analysis platform.
ASaiM is implemented as portable virtualized container based on Galaxy framework [8] .
Galaxy provides researchers with means to reproduce their own workflows analyses, rerun entire pipelines, or publish and share them with others. Based on Galaxy, ASaiM is scalable from single CPU installations to large multi-node high performance computing environments.
Deployments can be archived by using a pre-built ASaiM Docker image, which is based on the Galaxy Docker project ( http://bgruening.github.io/docker-galaxy-stable ) or by installing all needed components into an already existing Galaxy instance. This ASaiM Docker instance is customized with a variety of selected tools, workflows, Interactive tours and data that have been added as additional layers on top of the generic Galaxy Docker instance. The containerization keeps the deployment task to a minimum. The selected Galaxy tools are automatically installed from the Galaxy ToolShed [10] ( https://toolshed.g2.bx.psu.edu/ ) using the Galaxy API BioBlend [11] and the installation of the tools and their dependencies are automatically resolved using packages available through Bioconda ( https://bioconda.github.io ). We migrated then 10 tools/suites of tools and their dependencies to Bioconda ( e.g. HUMAnN2) and integrated 14 suites into Galaxy ( e.g. QIIME with around forty tools).
The containerization as well as the packaging with conda enables automatic continuous integration tests at different levels: dependencies (BioConda), tool integration in Galaxy, Galaxy itself and at ASaiM level. Together with strict version management on all levels, this contributes to a high degree of error-control and reproducibility.

Workflows
Each tool can be used separately in an explorative manner or multiple tools can be orchestrated inside workflows passing raw data to the data reduction step, to information extraction and visualization. To assist in microbiota analyses, several default but customizable workflows are proposed in ASaiM. All the available workflows with tool and parameter choices are documented ( http://asaim.readthedocs.org/ ).

Analysis of raw metagenomic or metatranscriptomic shotgun data
A workflow quickly produces, from raw metagenomic or metatranscriptomic shotgun data, accurate and precise taxonomic assignations, wide extended functional results and taxonomically related metabolism information (Figure 1). This workflow consists of i) processing with quality control/trimming (FastQC and Trim Galore!) and dereplication (VSearch [14] ; ii) taxonomic analyses with assignation (MetaPhlAn2 [12] ) and visualization

Analysis of amplicon data
To analyze amplicon data, the Mothur and QIIME tool suites are available to ASaiM. We integrated the workflows described in tutorials of Mothur and QIIME websites, as example of amplicon data analyses as well as support for the training material. These workflows, as any workflows available in ASaiM, can be adapted for a specific analysis or used as subworkflows by the users.

Running as in EBI metagenomics
The tools used in the EBI Metagenomics pipeline are also available in ASaiM. We integrate then also a workflow with the same steps as the EBI Metagenomics pipeline. Analyses made in EBI Metagenomics website can be then reproduced locally, without having to wait for availability of EBI Metagenomics or to upload any data on EBI Metagenomics. However the parameters must be defined by the user as we can not find them on EBI Metagenomics documentation.

Documentation and training
A tool or software is easier to use if it is well documented. Hence extensive documentation helps the users to be familiar with the tool and also prevents mis-usage. For ASaiM, we developed an extensive online documentation ( http://asaim.readthedocs.io/ ), mainly to explain how to use it, how to deploy it, which tools are integrated with small documentation about these tools, which workflows are integrated and how to use them.
In addition to this online documentation, Galaxy Interactive Tours are included inside the Galaxy instance. Such tours guide users through an entire analysis in an interactive (step-by-step) way. Some tours, included in every Galaxy instance, explains how to use Galaxy. We also developed such tours dedicated specifically to the ASaiM workflows. ASaiM is production-ready. It can also be configured to use external accessible computer clusters or cloud environments.
It is also possible and easy to install all or only a subset of tools of the ASaiM framework on existing Galaxy instances. The set of available tools can be easily extended either only a given instance using the Galaxy admin interface or for ASaiM more generally thanks to the simple definition of the installed tools in YAML files available in ASaiM GitHub repository. In the latter case, the Docker image will be automatically rebuilded and the already integrated tools will be updated to keep ASaiM up-to-date. For reproducibility reason, every version of the Docker image is associated to a tag and is conserved.

Conclusion
ASaiM provides a powerful framework to easily and quickly analyze microbiota data in a

Competing interests
The author(s) declare that they have no competing interests.

Funding
The Auvergne Regional Council and the European Regional Development Fund have