2.6.7 (mosga.py)

Upload your assembled FASTA eukaryotic genome file.

Needed for the GenBank/ENA submission files.
Assembly details for the submission (optional): GenBank needs all three as structured comment, ENA in the Webin-CLI manifest. MOSGA adds the annotation details itself.
Priority (highest priority first)
    Overlapping features
    How overlaps of these types are solved, only types of the selected tools are listed. Removed features are listed with the reason in removed.gff of the results. Overlapping rRNAs are always solved (table2asn rejects them).

    CpG island detection

    CpG island detection

    Splicing site detection

    Functional Enrichment Analysis

    Protein-Protein Interactions Analysis

    Protein-Protein Interactions Analysis

    Choose your tools:

    Genes

    Gene
    Protein-coding genes

    Prediction of gene locations and splice sites. Several predictors are combined into one gene set by EVidenceModeler.

    Mode of work

    Evidence-based or ab initio prediction.

    Combination of gene predictions

    Used if several gene predictors are selected.

    Functional Annotation

    Functional gene prediction.

    Repeats

    Repeats

    Detection of repeating sequences.

    tRNAs

    tRNA

    Prediction of tRNA sequences.

    rRNAs

    rRNA

    Search for rRNA sequence matches.

    ncRNAs

    ncRNA

    Search for non-coding RNA genes.

    Telomeres

    Telomere

    Detection of telomeric repeats.

    Assembly Validation

    Genome Completeness

    Validate genome completeness.

    Annotation Statistics

    Statistics of the final annotation.

    Quality-Control

    Contamination detection.

    UID Name Files Submission Date Start date End date Mode Status

    The Modular Open-Source Genome Annotator (MOSGA) is a pipeline that easily creates draft genome annotations through a graphical user interface. It combines several specific prediction tools and generates a submission-ready annotation file.

    The source code is freely available on GitLab and Zenodo (DOI: 10.5281/zenodo.5121228). We recommend building a new Docker container from the Dockerfile in the linked GitLab repository. MOSGA is modular and allows the easy integration of new prediction tools or even of whole third-party pipelines.

    No command line

    Upload a genome, choose the tools with a click and follow the job in the browser.

    Submission-ready

    NCBI .sqn/.tbl and ENA EMBL files with a Webin manifest; a submission check tells you what to fix.

    Modular tools

    Genes, repeats, tRNAs, rRNAs, ncRNAs and telomeres, freely combined from established predictors.

    Conflict resolution

    Overlaps are resolved by type and tool priority; removed features are listed with the reason.

    Functional annotation

    Product names from Swiss-Prot, eggNOG and IDTAXA; names to review are flagged.

    Quality control

    BUSCO, EukCC, AGAT, BlobTools and VecScreen in the same job.

    Comparative genomics

    ANI, gene comparison, phylogenetic trees and completeness across genomes, also of earlier jobs.

    Organelle scan

    Finds likely mitochondrial and plastid scaffolds and suggests submitting them as organelle sequences.

    Interactive results

    JBrowse 2 genome browser, summaries and all files on one results page.

    One Docker image

    Pinned and checksum-verified tool versions for reproducible results; self-hosted, so the data stays in-house.

    Convenient workflow

    Uploads of several GB in resumable parts, profiles per kind of job, "Submit again", retry of failed jobs and e-mail notifications.

    Open source

    MIT-licensed; the code is on GitLab and Zenodo, and new tools and merge requests are welcome.

    For any questions or comments, please contact us: roman.martin@uni-marburg.de. We are happy to receive new suggestions or even merge requests for a pipeline extension. For an overview of how MOSGA works, we recommend reading our GitLab wiki.

    We provide an example data set: the draft genome annotation of the Cafeteria roenbergensis BVI strain. We used an early version of MOSGA to annotate this genome (Hackl et al., 2020). Hackl, T., Martin, R., Barenhoff, K. et al. Four high-quality draft genome assemblies of the marine heterotrophic nanoflagellate Cafeteria roenbergensis. Sci Data 7, 29 (2020).

    We provide two examples of the comparative genomics workflow: the Saccharomyces species phylogenetics and the Saccharomyces gene comparison. An example annotation job for the organelle scanner, based on the Nannochloropsis oceanica genome, is available here.

    Please take note of the licenses of the selected tools.

    Whenever you use MOSGA, please cite us:
    Roman Martin orcid, Thomas Hackl orcid, Georges Hattab orcid, Matthias Fischer orcid, Dominik Heider orcid (2020). MOSGA: Modular Open-Source Genome Annotator. Bioinformatics. 36(22-23). 5514–5515. doi: 10.1093/bioinformatics/btaa1003.

    Roman Martin orcid, Hagen Dreßler orcid, Georges Hattab orcid, Thomas Hackl orcid, Matthias Fischer orcid, Dominik Heider orcid (2021). MOSGA 2: Comparative genomics and validation tools. Computational and Structural Biotechnology Journal. 19. 5504-5509. doi: 10.1016/j.csbj.2021.09.024.

    This page explains how to read the results of MOSGA. On the results page of a job, the buttons next to a section give a short explanation and link to the topic here.

    Jobs, queue and status

    After the submission, a job waits in the queue and runs as soon as a slot of the server is free. The status bar of the job shows its state (queued, running, finished, failed), the queue position and an estimate when it will be finished. The estimate is learned from the runtimes of former jobs on this server (the tools you chose, the genome size, the jobs ahead of yours), so it improves over time; it remains an estimate.

    Keep the link of the job (or register your e-mail address on the job page for a notification): results are stored for a limited time only, the job page shows until when. Submit again with these settings starts a new job with the same settings, e.g. with another tool or a corrected file.

    Result files and submission

    SQN Submission File (NCBI)
    The annotated genome for GenBank, created by NCBI's table2asn. Read the submission check before you submit it.
    EMBL Flat File and Webin-CLI Manifest (ENA)
    The same annotation for the European Nucleotide Archive; the manifest is the input of Webin-CLI.
    Organelles
    Sequences you marked as mitochondrion or plastid are submitted as own records with their own files.
    Genome Feature Table
    The annotation in the five-column feature table format of NCBI.
    View results online
    Opens the genome browser.

    The GFF3 of the final annotation and the predicted proteins are under Details. The submission files need details that only you know (organism, BioProject, assembly method): MOSGA fills in what was entered in the submission form.

    Genome browser

    JBrowse shows the genome with one track per tool and the final annotation of MOSGA. Tracks of different tools often disagree: MOSGA resolves overlaps by the priorities of the conflict resolution (submission form), the dropped features are listed in removed.gff with their reason. Compare the tool tracks with the final annotation to see which prediction was kept.

    Summaries

    Overall summary: statistics of the assembly — total length, number of scaffolds, N50/N90 (half or 90 % of the genome is in scaffolds at least this long; higher is more contiguous) and the GC content — and the number of features per type.

    Annotation summary: the features per sequence and type and the results of the quality checks (e.g. internal stretches of N). Things worth a closer look: very few or very many genes for the genome size, many single-exon genes in a species with introns, or genes on scaffolds that the organelle scan or VecScreen flagged.

    Details and removed features

    Genome Annotation GFF, Proteins (FASTA)
    The final annotation and the protein sequences of its genes.
    Removed Features (reasons)
    removed.gff lists every feature that is not in the final annotation, with the reason: an overlap lost by the conflict resolution (which feature won and why) or a filter (e.g. a too short protein). To keep such features, change the conflict resolution or the filters and submit again.
    Product Names to Review
    Protein names from the functional annotation that NCBI may not accept (e.g. "hypothetical", database IDs); check them before the submission.
    Single outputs
    The unprocessed output of every tool, for your own analyses.

    Submission check

    MOSGA runs the NCBI validation (table2asn) on the submission file and sorts every message by what to do:

    Needs your data
    Information only you can provide, e.g. the organism, the BioProject or assembly details. Add it in the submission details and submit again.
    Check before the submission
    Names or features that NCBI may question; check them.
    For information
    No action needed.
    Should not happen with MOSGA
    Please report it to us, or check the settings named in the hint.

    If an organelle sequence was submitted as nuclear DNA (e.g. a 16S rRNA on a scaffold), the page offers to submit the job again with it as organelle sequence.

    Validation

    The validation tools estimate how complete the assembly and the gene prediction are (BUSCO, EukCC) and whether the assembly contains DNA of other organisms (BlobTools, VecScreen). They do not change the annotation.

    BUSCO

    BUSCO searches genes that are present in a single copy in almost all species of a lineage. The result is given in percent of these genes:

    • C complete, split into S single-copy and D duplicated
    • F fragmented (only partly found)
    • M missing

    BUSCO (genome) measures the assembly, BUSCO (predicted proteins) the gene prediction. Good assemblies reach well above 90 % complete in a fitting lineage. If the proteins score clearly lower than the genome, the gene prediction missed genes: try another predictor, a closer species model or RNA-seq evidence. Many duplicated BUSCOs point to uncollapsed haplotypes or a polyploid genome; many fragmented ones to a fragmented assembly or incomplete gene models. Choose the most specific lineage of your organism.

    In the comparative mode, the BUSCO results of all genomes are compared; a genome with low completeness lacks genes and can distort the comparison and the tree.

    EukCC

    EukCC estimates the completeness and the contamination of a eukaryotic genome from lineage-specific marker genes. As a rule of thumb, aim for high completeness and a contamination of a few percent at most; higher contamination means that the assembly likely contains DNA of other organisms (check BlobTools). EukCC predicts the proteins with MetaEuk; it is meant for microbial eukaryotes (protists, fungi), its marker sets do not cover plants and animals well.

    BlobTools

    Every scaffold is a point placed by its GC content and its read coverage and colored by the taxon of its best database hit. Scaffolds of the target organism form a cloud; scaffolds of another taxon, or with a clearly different GC content or coverage, are candidates for contamination (bacteria, symbionts, food organisms). Check them and remove them from the assembly before the submission if they are foreign.

    Organelle scan

    The scan compares the scaffolds with mitochondrial and plastid proteins and lists the scaffolds with the most hits. A scaffold with many organelle genes is probably (part of) the mitochondrial or plastid genome. NCBI and ENA expect organelle genomes as own records: enter such scaffolds in the submission details as organelle sequences and submit again. The Mitos and plastid outputs show the organelle gene annotation.

    VecScreen

    VecScreen searches the assembly for vector, adapter and primer sequences (UniVec). Strong and moderate matches, especially at the ends of scaffolds, are typical leftovers of the sequencing and should be trimmed before the submission; NCBI checks this as well. Weak matches in the middle of a scaffold are often chance hits.

    Further modules

    CpG islands (CpGIScan, newcpgreport)
    Regions rich in CpG dinucleotides, often near promoters.
    Spliceator
    Predicted splice sites, a check of the intron borders of the gene models.
    g:Profiler, STRING, IID
    Functional enrichment and protein interaction networks of the annotated genes; they send gene names to these external services and work best for organisms these databases know.
    Log Viewer, Snakemake DAG
    The logs of all steps and the graph of the workflow of the job.

    Phylogenetic tree

    The tree is computed from conserved single-copy genes (BUSCO) shared by the genomes: they are aligned, trimmed and used for a maximum-likelihood tree (Newick file and figure). Branch lengths are substitutions per site. A genome with low completeness contributes few genes and is placed less reliably. Add an outgroup to root the tree.

    Average nucleotide identity

    ANI is the mean identity of the regions two genomes share: 100 % for identical genomes, lower for more distant ones. The heatmap groups similar genomes. For prokaryotes, about 95 % ANI is used as species border; for eukaryotes there is no fixed border, use the values to compare the genomes with each other.

    Gene comparison

    The protein-coding genes of the genomes are searched in each other. The table lists which genes have a match in which genome above the identity threshold, the figure summarizes how similar the genomes are by their shared genes. Genes without a match are candidates for lineage-specific genes — or for genes the prediction missed in the other genome.

    Failed jobs

    If a step fails, the job stops and keeps the steps that finished. Error Identification names the failing step and the most likely reason. The Log Viewer in the sidebar opens run.txt first: the end of it shows the failing rule; its own log (named in the error) has the details. Common reasons are a wrong file format (e.g. a FASTQ where a hints file is expected) or a genome that is too large for a tool. Retry starts the job again from the failed step; if the reason was your input, use Submit again with corrected files. If you cannot find the reason, please contact us with the job link.

    Citing

    The results page lists the publications of all tools your job used (What to cite under Details). Please cite them together with MOSGA, see the About tab.

    The University of Münster hosts this MOSGA instance for demonstration purposes.
    It runs on a virtualized AMD Ryzen CPU with 8 cores and 32 GB of memory.

    We keep the 100 most recent job submissions online. Once this limit is exceeded, we delete the oldest job that is more than 14 days old. Incoming jobs are queued and processed as soon as possible. Jobs that occupy our hardware for more than 48 hours may be terminated. We recommend not uploading files larger than 2 GiB.

    We reserve the right to analyze failed jobs in order to identify errors and to provide bug fixes and quality improvements. Even so, your results will not be shared with anyone and will be deleted regularly.

    If you provide a notification email address, we may contact you if your job fails so that we can help resolve the issue.

    Imprint

    Universität Münster
    Medizinische Fakultät (Faculty of Medicine)
    Albert-Schweitzer-Campus 1, Building D3
    48149 Münster, Germany
    Phone: +49 251 83 52263
    Fax: +49 251 83 55004
    Email: dekanmed@ukmuenster.de

    The Faculty of Medicine is part of the University of Münster. The University of Münster is a corporation under public law (Körperschaft des öffentlichen Rechts) and also an institution of the State of North Rhine-Westphalia. It is represented by its Rector, Prof. Dr. Susanne Menzel-Riedl.

    Supervisory authority:
    Ministerium für Kultur und Wissenschaft des Landes Nordrhein-Westfalen
    Völklinger Straße 49
    40221 Düsseldorf, Germany

    VAT ID: DE 126118759

    Responsible for content pursuant to § 18 (2) MStV:
    The Dean of the Faculty of Medicine
    Prof. Dr. Frank Ulrich Müller
    Albert-Schweitzer-Campus 1, Building D3
    48149 Münster, Germany
    Phone: +49 251 83 52263
    Fax: +49 251 83 55004
    Email: dekanmed@ukmuenster.de


    Server usage