A major goal during the development of the MGP Portal was to create a reproducible workflow for the creation of well-annotated and visually accessible genome repositories (Figure 1). To that end, we have adopted the JavaScript-based JBrowse 18 as the engine for our Mnemiopsis Genome Browser, resulting in a clean and responsive user interface for viewing the genome assembly, gene models, and all supporting data. We have developed a number of tools in the Perl scripting language to convert sequence and annotation data into a format accepted by JBrowse (version 1.11.1), which are included as Additional files 1, 2 and 3. In addition, we provide a Perl script that facilitates the creation of MediaWiki pages that display nucleotide and protein sequences, exonic genomic coordinates, and PFAM domains (Additional file 4). The actual creation of genome assemblies and annotation files is outside the scope of this manuscript but is outlined in detail elsewhere 4 .

To build feature tracks, JBrowse requires properly formatted input files. While newer versions of JBrowse released subsequent to the creation of the Mnemiopsis Genome Browser are able to accept a slightly expanded number of inputs (including specific relational database dumps), the generic feature format version 3 (GFF3) flat file was the input type we adopted, a format that continues to be supported by JBrowse at the time of this writing (version 1.11.1). Adoption of the GFF3 format greatly facilitated the processing of multiple output file types produced by a number of data analysis programs. To begin, the scaffoldToGFF3.pl script (Additional file 1) can be used to reformat scaffold sequences from FASTA to GFF3. Two parameters (–i and –o) are required, specifying the input scaffold file and desired output directory, respectively. Two optional parameters (–l and –n) are also available. The first should be called if the gap-indication letter in the input is different from ‘N’ , accepting a single character (e.g., X) as a substitute, while the second parameter can be called in order to specify which feature-generating program was used. The script creates a GFF3 file for each sequence in the input scaffold FASTA file and is able to handle both gapped [e.g., the scaffold (SCF) track] and repeat-masked [e.g., the repeat-masked (MASK) track] regions in a scaffold. Next, evmToGFF3.pl (Additional file 2) parses a GFF3-formatted output file created by EvidenceModeler (EVM) 19 by collecting data about the start and end positions of predicted genes, using this information to create well-formed GFF3 files. The script accepts several additional parameters; the input EVM file (–i) and desired output directory (–o) must be set, while the third and optional parameter (–n) is, again, the name of the feature-generating program (e.g., EVM). The third module is called cufflinksToGFF3.pl (Additional file 3) and is used to parse predicted transcript assemblies from the GTF-formatted file created by Cufflinks 20 . The cufflinksToGFF3.pl script has the same behavior for transcript location as evmToGFF3.pl has for predicted gene location, and accepts the same three parameters (–i, –o, and –n).

To import GFF3 data into JBrowse for display as custom tracks in the main genome window, a series of three JBrowse-supplied Perl scripts (prepare-refseqs.pl, biodb-to-json.pl, and generate-names.pl) need to be executed using appropriate parameters and a system-specific configuration file, the details of which can be found in the tutorials on the JBrowse Web site (http://www.jbrowse.org). A number of custom CGI scripts were written to create the hyperlinks connecting features in JBrowse to the various sources of gene data.

A Perl script named create_wiki_page.pl has been used to create MediaWiki pages for displaying genomic data (Additional file 4). Here, we provide a sample wiki page that takes FASTA-formatted nucleotide and protein sequences, GFF3 files containing exonic coordinates, and an hmmscan output file containing information on Pfam-A domains as input. The create_wiki_page.pl script requires five parameters. The first three parameters specify a set of input files, including a nucleotide FASTA file (-n), a protein FASTA file (-a), and a Pfam-A file (-p). The remaining parameters specify a directory containing the input GFF3 files containing the exonic coordinates (-d) and an output directory (-o). In addition, we show a PHP command line that imports a wiki page into MediaWiki:php/$WIKIHOME/maintenance/importTextFile.php - -user=USER wikipage.out The wikipage.out file is created by the create_wiki_page.pl.

Researchers interested in creating a customized genome browser or gene wiki for visualizing genome-related data are encouraged to utilize the aforementioned scripts (Additional files 1, 2, 3 and 4), as their implementation satisfies the fundamental requirements of both JBrowse and MediaWiki. Furthermore, these modules may serve as a useful framework for both the development of gene wikis and more advanced genome browser tracks as new JBrowse applications are created to visualize genomic data.

Mnemiopsis sequences are stored as individual text files, and several Perl scripts were written to retrieve these as single combined files. Multiple sequences can be downloaded via an HTML interface that was developed using a series of CGI/Perl scripts. The Mnemiopsis BLAST tool was implemented using ViroBLAST 21 and runs on an Apache Web server (Additional file 5: Figure S1). The ViroBLAST tool was written in PHP, a server-side scripting language 22 , and Perl 23 , applying the stand-alone blastall program downloaded from NCBI 24 . The Mnemiopsis BLAST databases were created using formatdb. PHP was used to parse the BLAST output created by ViroBLAST to generate the formatted BLAST results, including the internal links to the Genome Browser, the Gene Wiki pages and the Fetch Scaffold Tool. MediaWiki (version 1.19.11), written in PHP and Perl, was used for the Gene Wiki implementation. Perl was also used to create data for displaying on the Gene Wiki pages. The KEGG pathways pages were developed using JavaScript to display KEGG identifier, pathway, and gene symbol search lists. CGI/Perl scripts were used for KEGG pathway search functions and data downloading utilities (Additional file 6: Figure S2). Python (version 2.6) 25 scripts were used to search the Pfam domains, identified using hmmscan 26 from the HMMER suite, and CGI and JavaScript were used to display the Pfam domain search results (Additional file 7: Figure S3).

One feature of the Mnemiopsis Genome Project Portal is a customized stand-alone Web-based BLAST interface for performing nucleotide and amino acid sequence similarity searches (Additional file 5: Figure S1). ViroBLAST was used to implement our Mnemiopsis BLAST tool, producing an organized, manageable output that is easy to parse and navigate. Users may input their FASTA-formatted query sequences directly into the search box or upload sequence files from their computer. The customary set of BLAST programs is available, including BLASTN, BLASTP, BLASTX, TBLASTN, and TBLASTX. Nucleotide sequence databases include the Mnemiopsis genomic scaffolds (Main Scaffolds), consensus gene prediction models (Gene Models 2.2) and Unfiltered Gene Models (unincorporated predictions) described in Ryan et al. 4 , all publicly available Mnemiopsis ESTs and mRNAs from GenBank (Public ESTs), the Mnemiopsis mitochondrial genome 27 , Cufflinks-assembled RNA-seq transcripts, and Trinity-assembled RNA-seq transcripts 28 . The protein sequence databases available through the MGP Portal include the translated proteins derived from the Mnemiopsis consensus gene prediction models (Protein Models 2.2), the unincorporated Mnemiopsis proteins derived from unincorporated gene prediction models (Unfiltered Protein Models), and the computationally derived Mnemiopsis mitochondrial proteins. Additionally, a user may create a customized user-defined BLAST database by uploading a file containing FASTA-formatted nucleotide or protein sequences of interest. BLAST output results feature customized color-coded boxes linked directly to relevant internal annotation resources, including the Mnemiopsis Genome Browser [B], the wiki-based Mnemiopsis Gene Pages [G], the Scaffold Fetch Tool [S], Unfiltered Gene Models [U], Cufflinks-assembled transcripts [C], and Trinity-assembled transcripts [T] (Figure 2).

One of our primary objectives in developing the MGP Portal is to provide a graphical tool for the scientific community to visualize the various types of Mnemiopsis genome data currently available. Using the built-in JBrowse genome browser, users can view a variety of data tracks, such as the Mnemiopsis genome assembly, gene prediction models, and RNA-seq data (Figure 3). Several customized JBrowse tracks are available for viewing, including consensus gene prediction models (labeled 2.2 in the browser), Mnemiopsis RNA-seq reads assembled into transcripts using Cufflinks (CL2) and Trinity (TRN15-30 hpf), publicly available Mnemiopsis ESTs from GenBank (EST), publicly available Mnemiopsis mRNAs from GenBank (GBNT), assembled genomic scaffolds (SCF), genomic regions that have been repeat-masked using V-Match (MASK) 29 , experimentally verified Mnemiopsis RACE PCR transcripts (RACE), unincorporated gene prediction models (2.2UF), and non-redundant protein domains derived from Pfam hmmscan runs using the 2.2 and 2.2UF datasets and the six-frame translations of the Mnemiopsis genome (PFAM2.2).

The JBrowse display is organized by genomic scaffold. Scaffolds are named using a six-character convention (e.g., MLnnnn); the ML designates the species (Mnemiopsis leidyi), and the individual scaffolds are numbered from 0001 to 5100 (e.g., ML0001). Gene identifiers (e.g., ML000129a) start with a prefix indicating the scaffold on which the gene is located (in this example, ‘ML0001’), followed by a non-padded integer that is unique in combination with the scaffold identifier (in this case, ‘29’) and ending with a lower-case letter that specifies the gene isoform (in this case, ‘a’). Data displayed in the browser can be searched using a variety of Mnemiopsis identifiers. A scaffold-based query takes the user directly to that scaffold, while a gene-based search goes to that gene’s location on the appropriate scaffold. A user may also search the genome browser using a Mnemiopsis GenBank mRNA identifier, an EST identifier, or a Pfam-A domain name (e.g., AF293700.1, FC475136, or “Glyco hydro 20”, respectively). The PFAM2.2 track was created by running hmmscan against the Protein Models 2.2, the Unfiltered Protein Models, and the six-frame translations of the Mnemiopsis genome. Scaffold coordinates are displayed across the top of the browser window. Navigation options can be found beneath the coordinate bar, including the zoom tool and left-right arrows. Users can also refine the displayed region by entering the scaffold coordinates into the search box to the right of the navigation options.

Genome browser tracks are described in the Track Descriptions link above the left sidebar. The consensus Mnemiopsis gene prediction models (track 2.2) are presented by default. Additional track can be added by clicking on the appropriate track box on the left sidebar of the main view window. All track options are displayed in a given scaffold window even when there are no annotated features in that particular region. Features are represented as black arrows, with the direction of the arrow indicating the orientation. Exons are presented as light-colored solid bars (e.g., green for 2.2 and pink for 2.2UF), while untranslated regions (both 5’ and 3’) are rendered using darker-shaded colors. Assembled genomic scaffolds (SCF) are depicted as solid black tracks with intermittent bright pink gaps. The MASK track appears as a solid black bar, with blue highlighting the regions that were repeat-masked using VMatch (Figure 3). The Reference sequence track depicts the scaffold sequence, but only when the display is fully zoomed in.

Previously, we identified gene clusters that contain likely orthologs of both Mnemiopsis and human proteins 4 . In order to gain some insight into the function of individual Mnemiopsis genes, we used this information to assign individual Mnemiopsis genes to human KEGG pathways. We converted Ensembl identifiers for all human genes from our phylogenetically informed clusters of orthologous genes 4 to Entrez Gene IDs using a file from the Entrez Gene FTP site 30 ; we then used EASE 31 to link Entrez Gene IDs to KEGG pathways. These data can be accessed by following the KEGG Pathways link found in the left sidebar of most MGP Portal pages. Human KEGG pathways containing genes with a Mnemiopsis ortholog are searchable by selecting a KEGG identifier (e.g., hsa00604), a pathway name (e.g., glycosphingolipid biosynthesis), or a gene symbol (e.g., SLC33A1) from their respective lists, or by using the KEGG pathways search box (Additional file 6: Figure S2). The results are presented as pathway-specific ortholog cluster matrices (Figure 4).

For each KEGG pathway, each row in the ortholog cluster table represents a cluster of orthologous genes from our clustering analysis 4 . For a cluster to be included in the table, at least one human gene from that cluster must be present in the KEGG pathway represented. In most cases, a row consists of one or more human genes that belong to the selected KEGG pathway, along with the computationally predicted orthologs from Mnemiopsis and 21 other model organisms. The ‘Cluster’ column indicates the most inclusive phylogenetic clade that encompasses all of the genes in the particular cluster (e.g., ‘Metazoa’ could indicate that the cluster contains genes from both bilaterians and cnidarians and thus cannot be characterized by a less inclusive clade, such as ‘Bilateria’). Each human gene is hyperlinked to its Entrez Gene entry. The ‘Ratio’ column represents the number of human genes in the particular cluster that are in a given pathway (numerator) over the number of total human genes in that cluster (denominator). The higher the ratio, the more likely the non-human orthologs in the cluster are involved in the pathway. The numbers in the columns under each species abbreviation indicate the number of genes from that species that are in that cluster. Each number in the ‘Ml’ (Mnemiopsis leidyi) column is linked to the appropriate Mnemiopsis gene ID(s) and corresponding Gene Wiki pages.

Another way to characterize the Mnemiopsis genes is to determine the protein domains that they encode. We used hmmscan from the HMMER suite (HMMER 3.0; March 2010) to search the Protein Models 2.2 and Unfiltered Protein Models for domains from the Pfam-A database (version 25). The gathering threshold (cut_ga) option was used to ensure conservative domain prediction. The Pfam Domains link on the home page of the MGP Portal takes the user to a query page, where researchers can specify a given Pfam-A domain by name or accession number, then search for Mnemiopsis genes that contain that domain (Additional file 7: Figure S3). The results are displayed as a list of protein models, listed by gene identifier, and the number of query domains found in those protein models. Additionally, a user may download FASTA-formatted Pfam-A domain sequences from the resulting list by clicking on the check boxes next to the sequence(s) of interest, selecting either Pfam-A domain only or the full-length domain-containing protein from the pull-down menu, and clicking ‘Get’.

In an effort to engage the collective expertise of the scientific community, we have implemented a collaborative wiki (MediaWiki version 1.19.11) for the Mnemiopsis gene complement. The Mnemiopsis Gene Wiki is accessible from the left sidebar of most pages and is searchable either by selecting a Mnemiopsis gene identifier (e.g., ML00011a) from the drop-down menu or by manually entering an identifier in the appropriate search box. Users can also access these pages by clicking on a gene in the 2.2 track of the genome browser. Each record in the Gene Wiki represents a single Mnemiopsis gene and provides the following annotation: nucleotide and protein sequences, coding exonic genomic coordinates, pre-computed BLAST hits from numerous organisms displaying the top hits for each protein, the top non-self BLAST hit to Mnemiopsis, Pfam-A domains, Gene Ontology (GO) functional annotations, human disease genes from Online Mendelian Inheritance in Man (OMIM), and a table of ortholog clusters formed by phylogenetically informed clustering methods 4 (Figure 5). In addition, controlled editable sections have been included that permit (and encourage) the scientific community to provide further gene annotation for isoforms, in situ images, references, and other notes for each gene. Users interested in supplementing our gene model annotation at the Mnemiopsis Gene Wiki pages must first create an account and log in prior to submitting their contributions. In-house subject matter expert data curators are notified by e-mail following the creation of a new user account or an edit to an existing Gene Wiki record. Any content changes or additions to the Gene Wiki are thoroughly evaluated by these data curators and are made public subject to their approval.

Pre-compiled BLAST hits are enumerated in tabular form. Each Mnemiopsis protein was compared to the UniProt and NCBI non-redundant protein databases (nr) using BLASTP. The results display the hit number, the accession numbers, E-values, and brief descriptions of the top four hits (lowest E-values). Accession numbers are linked to relevant corresponding entries at UniProt and GenBank. The E-values are hyperlinked to the pairwise BLAST alignments.

Each Mnemiopsis protein was also compared to sequence data from developmentally relevant organisms, including Homo sapiens, Drosophila melanogaster, Capitella teleta, Amphimedon queenslandica, Nematostella vectensis, Hydra magnipapillata, Trichoplax adhaerens, Monosiga brevicollis, Salpingoeca rosetta, Capsaspora owczarzaki, fungi, plants, and non-eukaryotes. The top hits for BLASTP and TBLASTN results, falling below an established E-value threshold (E-value ≤ 1 × 10^-6), are displayed along with their gene or protein identifiers, E-values, and description of the best hits. For the Mnemiopsis organismal database search, the gene identifier of the top non-self hit is displayed (and linked to its corresponding Gene Wiki page) along with the E-value for that alignment.

The Gene Wiki also contains a section displaying the Pfam-A domains that are encoded by the protein. The Pfam identifier, domain architectures, sequence start and end coordinates, HMM start and end coordinates, E-values, and domain sequence are displayed for each Pfam-A domain. All Pfam identifiers are hyperlinked to their corresponding entries on the Pfam Web site 32 . To further assist in classification, GO terms are presented for each gene, with GO terms assigned using the Argot2 33 method. Functional annotations derived using Blast2GO 34 are also presented for each gene.

The genomic sequence of a single (or partial) scaffold can be retrieved using the Fetch Scaffold Tool. A user can download a FASTA-formatted full genomic scaffold sequence by following the Fetch a Scaffold link in the left sidebar of the MGP Portal home page, entering a scaffold identifier (e.g., ML0001) in the search box, and selecting the Fetch sequence option (Figure 6). Partial scaffold sequences can be retrieved by adding scaffold coordinates to the above query. Alternatively, users can retrieve the reverse complement or six-frame protein translation of the scaffold by selecting the appropriate option.

As noted above, one of our primary objectives in developing the MGP Portal is to simplify the dissemination of all Mnemiopsis sequence data to the scientific community at large. To that end, we have provided users a direct method for obtaining entire Mnemiopsis datasets as compressed text files. The following complete sequence datasets can be downloaded by clicking on the appropriate links located in the left sidebar: the Mnemiopsis leidyi genome assembly (5,100 scaffolds), the full set of Mnemiopsis Gene Models (16,548 genes), the Mnemiopsis Protein Models (16,548 proteins), the Mnemiopsis Unfiltered Protein Models (60,006 proteins), all publicly available Mnemiopsis EST sequences from NCBI (15,752 ESTs), and the full Mnemiopsis mitochondrial genome and protein sequences 27 (11 proteins; Table 1). Optionally, users may enter a known Mnemiopsis sequence identifier (e.g., ML0001 or ML00011a) in the search box or select from a list of identifiers to retrieve a single scaffold, gene model, protein model, or EST of interest.

The MGP Portal was developed to facilitate research that would benefit from the availability of genomic information from this emerging model organism and, to this end, it includes a number of intuitive data analysis tools. To illustrate this point, consider the case of a developmental biologist studying the human TALE class homeobox gene family (e.g., PBX3; [GenBank:NP_001128250.1]) who may be interested in comparing these sequences against (or predicting novel) Mnemiopsis homeodomain orthologs. A straightforward approach to addressing this question would be to run a BLASTP search of the PBX3 protein sequence against the Mnemiopsis Protein Models (2.2) database. The Mnemiopsis BLAST results display a number of high-scoring candidate proteins that can be further evaluated for properties characteristic of TALE class homeodomains (e.g., a TALE-type homeobox; Figure 1).

Alternatively, another biologist may be interested in searching for novel Mnemiopsis homeodomains, using sequence data from a closely related organism such as Amphimedon as the basis for their search. One approach would be to retrieve a complete set of known Amphimedon homeodomain proteins through an NCBI Entrez query (Search < Protein > for: “homeodomain AND Amphimedon[ORGN]”, which yields 31 known Amphimedon homeodomain proteins at the time of this writing). FASTA sequences can then be copied and pasted into the Mnemiopsis BLAST search window or uploaded as a file from a local computer. A unique feature of the MGP Portal BLAST implementation includes access to the Unfiltered Protein Models database, which contains the complete unincorporated (unfiltered) protein dataset derived from the Mnemiopsis gene prediction and annotation process. A BLASTP search against the Unfiltered Protein Models database provides additional information about possible alternate transcripts and isoforms that were screened and filtered out during the initial annotation process. These unfiltered protein models can be placed into genomic context by following the color-coded Mnemiopsis Genome Browser [B] links in the far right-hand column of the BLAST results. The browser provides direct access to additional annotation, including a graphical representation of the Pfam-A domain that overlaps the protein model, as well as links to the sequence of the Pfam-A domain (Figure 7).

Similarly, a researcher may also want to use their own custom scripts or external computational tools to further-explore the available Mnemiopsis data sets. In such a case, the Download Sequence links in the MGP Portal can be used to download both the Protein Models and the Unfiltered Protein Models for analysis with tools from the HMMER suite 26 (e.g., hmmsearch Homeobox.hmm ML2.2aa > ML_novel_HDs). Predicted domains with E-values below an inclusion threshold (e.g., E-value < 0.05) could then be considered candidate homeodomains for further evaluation.