The genome portal of the Department of Energy Joint Genome Institute: 2014 updates

1A leader in genome sequencing, the DOE JGI has significantly expanded its capabilities to achieve a deeper understanding of biological functions encoded by DNA. With the increasingly diverse data types it generates and associated computational strategies for their analysis, the DOE JGI is focused on enabling its community of users to make important new scientific discoveries that address major energy and environmental challenges (U.S. Department of Energy Joint Genome Institute Progress Report, http://www.jgi.doe.gov/whoweare/JGI-progress-report-2012.pdf).

2In 2012, JGI completed 2635 projects, a 3-fold increase over 2011, and generated >56 trillion nucleotides of genome-sequence data from microbes and microbial communities, fungi, algae and plants. In the past year alone, JGI has added 650 genomes to the public databases. JGI sequencing efforts are mostly based on the proposals submitted to JGI by researchers and to a large extent are aligned with several large-scale JGI programs such as Community Science Program (CSP) and Genomic Encyclopedia of Bacteria and Archaea (GEBA). The amount of data being produced by sequencing and analyzing genomes brings genomics into the realm of big data category. The increased scale in data has demanded a corresponding advance in the scale and robustness of the computing infrastructure underlying the Genome Portal. Since the previous publication on the JGI Genome Portal in January 2012, we specifically focused on improvement of computational resources for efficient storage, access, download and analysis of these data.

3The Portal home page, available at http://genome.jgi.doe.gov, allows a user to select the genomic data set and the tools to work with (Figure 1). It provides the information about the latest genome releases and access to several specialized JGI database resources: integrated microbial genomes (IMG) and metagenomes (IMG/M), Phytozome for green plant genomes and MycoCosm for fungal genomes. It also provides access to extensive help pages as well as a recently developed interactive tutorial for novice users.

4There are two major categories of data available in JGI sequencing projects and annotated genomes and metagenomes. More than 15 000 DOE JGI projects of different types are publicly available and searchable in our database. These projects include different genomic products, such as standard and improved draft, finished genomes, gene expression profiling, resequencing, metagenome projects, single cell projects, transcriptomes, metatranscriptomes, exomes, specialized analysis types and others. The Genome Portal provides access to over 17 000 annotated genomes and metagenomes available in the DOE JGI database, along with specialized analytical tools to navigate these data sets and compare genomes.

5Individual Genome Portal project sites (‘portals’) are created automatically on passage of sample quality control stage for a given sequencing project. The JGI has ∼24 000 portals for isolate, metagenome, single-cell, transcriptome and resequencing projects. Portals aggregate relevant analysis tools for each project and provide download access to project data and project information such as proposal name, organism and taxonomy info as well as the names of principal investigator and other contacts relevant to an associated data set (Figure 2).

6To navigate the large volume and diversity of genome projects at the JGI, the Genome Portal offers several new and improved tools to locate a genome of interest: a list of all JGI projects, a ‘Search’ function, an interactive ‘Tree of Life’ and domain-specific comparative resources. Data access and the spectrum of data depend on the stage of each project and the role of users in these projects (e.g. administrative). Free registration allows users not only to access all data but also to initiate new projects with JGI.

7The ‘Project List’ link on the Genome Portal page (http://genome.jgi.doe.gov) (Figure 1) brings users to a list of all DOE JGI projects with a detailed description of each project including its scope and current status, taxon, the JGI program, the project lead and the computational resources available for this project.

8The JGI Genome Portal now includes a new Portal Search function that enables searching for genomes and projects by keyword (e.g. plants, algae, single cell, water), name and other categories of data (shown in a pull-down menu, Figure 1). The user can also locate projects related to specific classifications such as a proposal or a project. Typing the beginning of the word in the text window brings up a pull-down menu with relevant search term choices.

9Search results now include Project information and status, GenBank accession, taxonomic information, PI contact and much more (Figure 3). Every JGI Proposal, which can include multiple genomes, transcriptomes or other data sets, now has all related portals grouped together, and downloads of entire segments of data are possible at the proposal level.

10The Tree of Life organizes the sequenced genomes by the domains of life and metagenomes by ecological ‘niche’. Clicking on a branch name produces a menu displaying available genomes in the corresponding kingdom, phylum, class or order. Selecting an entire group of genomes or a particular genome connects a user to a corresponding organism page or pages in specialized resources indicated by associated icons (Figure 1). Microbial, archaeal genomes and metagenomes are directed to the corresponding IMG and IMG/M pages, fungal genomes to Mycocosm and plant genomes to Phytozome. Genomes not included in either of these specialized resources are linked to their individual portal pages containing information about the corresponding project and available tools.

11The Genome Portal Web site is built on Apache HTTPD, Tomcat and MySQL. A majority of the Genome Portal components have been developed using Java and a variety of available open-sources tools and technologies. There are four load-balanced web servers, talking to two back-end database servers. A web-driven automated build system that uses Jenkins (http://jenkins-ci.org/) takes each machine silently out of the cluster, builds a new version of the portal and puts the machine back into the cluster. This procedure ensures that updates can be applied without disruption to users. This setup further makes the portal resilient against hardware failures.

12The Genome Portal acquires data from several of JGI’s annotation pipelines via an API and it makes the data available to authorized users without delays. There is an advanced monitoring system in place that allows preventing problems that may impact Web site and database performance.

13Over the last 2 years, we have forged a strong alliance in high-performance computing with the National Energy Research Scientific Computing Center (NERSC). NERSC now hosts the servers on which the Genome Portal is running and provides access to ESnet (Energy Sciences Network), which allows for top of the class transfer rates.

14The Genome Portal currently supports download of multiple file types: all raw and quality control data, assemblies and various analysis files. All files can be accessed in a variety of ways, each designed with a particular class of users in mind.

15In the web-driven approach to data access, the user finds a portal of interest and clicks on the download tab of that portal. Downloads are available in a tree structure that divides the files into logical groups so that the user can download raw data files, assemblies and so on with a single operation (Figure 4).

16For users who need to download a large number of files or large files (with the largest being several hundred gigabytes), data are available via the public JGI endpoint on GlobusOnline (https://www.globusonline.org/). This requires that the user log on to globusonline.org and activate the ‘jgi#portal’ endpoint using their JGI credentials. GlobusOnline then provides a user-friendly interface for facilitating big data transfers. Behind the scenes the transfers are performed using GridFTP, which is a parallel transfer protocol and program. GridFTP has built-in checks that ensure the integrity of the transfers and guarantees that the files reach their destination intact.

17Finally, a web service is available for users who seek to download JGI data programmatically. This service is described in the documentation on the Genome Portal Web site.

18Whether a user navigates the web user interface, uses GlobusOnline or uses the web service, the data that can be accessed are the same in each case. As the amount of data the JGI makes available is counted in petabytes, many of the files are stored on the NERSC HPSS mass storage archive. All three approaches hide this fact from the user, interacting with the tape system to cache files as required.

19In addition to downloading data, The Department of Energy Systems Biology Knowledgebase (KBase) (http://kbase.science.energy.gov/) users can upload their data directly from the portal page to KBase. They do so by selecting files in the tree on the download page, clicking ‘Push to KBase’ and entering their KBase credentials.

20The increasing volume of genome data has led the NCBI to develop a new automatic submission system to streamline the process of submitting complex high-throughput sequencing or functional genomic data sets. The DOE JGI has collaborated with the NCBI to develop the first major implementation for a sequencing center of a framework to use this submission system. Using this system, the JGI Genome Portal automatically generates and monitors BioSample and BioProject submissions to NCBI using an XML-based submission protocol. Internally, we use the same infrastructure to prepare GenBank submissions. The latter are submitted to the NCBI FTP server using the ASN.1 and related GenBank-specific formats.