GFF3-to-DDBJ

日本語版はこちら。

Overview

GFF3-to-DDBJ converts GFF3 and FASTA files into the DDBJ annotation format required for submission. It is the DDBJ-specific equivalent of tools like table2asn (NCBI) or EMBLmyGFF3 (ENA).

View the tests/golden directory for example output (.ann files).

Accuracy and Validation

Since “perfect” GFF3-to-DDBJ conversion is not formally defined, this tool uses RefSeq GFF3-GenBank correspondence as a gold standard. We validate output by:

1. Comparing gff3-to-ddbj results against GenBank-sourced annotations via an internal genbank-to-ddbj tool.
2. Passing all output through the DDBJ BioProject/BioSample/Sequence Data (MSS) Parser.

Installation

Via Bioconda

conda create -n ddbj -c conda-forge -c bioconda gff3toddbj
conda activate ddbj

Via PyPI

conda create -n ddbj -c conda-forge -c bioconda pip samtools
conda activate ddbj
python -m pip install gff3toddbj

Via GitHub (Nightly)

conda create -n ddbj pip
conda activate ddbj
python -m pip install 'git+https://github.com/yamaton/gff3toddbj'

Usage

gff3-to-ddbj \
  --fasta myfile.fa \               # Required
  --gff3 myfile.gff3 \              # Strongly Recommended (bare-minimum if absent)
  --metadata mymetadata.toml \      # Optional
  --locus_tag_prefix PREFIX_ \      # Required for BioSample
  --transl_table 1 \                # Default: 1 (Standard)
  --output output.ann               # Optional: stdout by default

Argument Details

• --locus_tag_prefix: The prefix assigned by BioSample.
• --transl_table: Genetic code index (e.g., 11 for Bacteria). See DDBJ Genetic Codes.

Under the Hood

GFF3-to-DDBJ processes your data through the following pipeline:

1. Data Preparation

• FASTA Compression: If the input is standard Gzip, the tool re-compresses it using bgzip (e.g., creating myfile_bgzip.fa.gz). This enables indexing and reduces memory usage; the resulting file remains compatible with standard gzip tools.
• Gap Detection: Scans FASTA sequences for N runs and automatically generates assembly_gap features.
• Topology Handling: If GFF3 has Is_circular=true, the tool inserts a TOPOLOGY feature and manages origin-spanning features.

2. Feature & Qualifier Mapping

• SO-to-INSDC Translation: Maps GFF3 “types” to DDBJ “Features” based on Sequence Ontology.
  • Example: transcript (SO:0000673) is translated to a misc_RNA feature.
• Qualifier Renaming: Converts GFF3 attributes to DDBJ-compliant qualifiers based on renaming rules.
  • Example: ID=foobar becomes /note="ID:foobar".
• Genetic Code Assignment: Automatically adds the /transl_table qualifier to every CDS feature based on the user-provided index (default: 1).

3. Coordinate Processing

• Joining: Features sharing a parent are merged using join() notation. This applies to CDS, exon, mat_peptide, V_segment, C_region, D-loop, and misc_feature.
• RNA/Exon Logic: The location of joined exons is assigned to the parent RNA’s location, and individual exon entries are discarded.
  • Note: exons are not joined if their direct parent is a gene.
• Partialness: Adds partial indicators (< or >) to CDS locations if start or stop codons are missing. (See: Offset of the frame at translation initiation by codon_start).

4. DDBJ Compliance Logic (Product & Gene)

• Product Enforcement: To conform to DDBJ instructions, each CDS is restricted to a single /product:
  • Even if there are multiple general names for the same product, do not enter multiple names in ‘product’. Do not use needless symbolic letters as delimiter for multiple names. If you would like to describe more than two names, please enter one of the most representative name in /product qualifier, and other(s) in /note qualifier.
  • If the name and function are not known, we recommend to describe as “hypothetical protein”.
• Gene Consistency:
  • Ensures the /gene qualifier has a single value; additional values move to /gene_synonym. (Reference: Definition of Qualifier key: /gene).
  • Copies /gene and /gene_synonym qualifiers from parent gene features to all children (e.g., mRNA, CDS).

5. Metadata & Filtering

• Metadata Injection: Inserts source information and global qualifiers from the metadata file. See “Metadata Configuration” in “Customization” below.
• Compliance Filtering: Removes features and qualifiers violating the DDBJ usage matrix.
  • Note: The gene feature is discarded by default in this process.
• Deduplication: Removes redundant qualifier values generated during processing.

6. Final Formatting

• Sorting: Lines are ordered by start position, feature priority (placing source and TOPOLOGY at the top), and end position.

• Validation Logs: Displays all discarded items via stderr:

  WARNING: [Discarded] feature -------> gene (count: 49911)
  WARNING: [Discarded] (Feature, Qualifier) = (mRNA, Parent) (count: 57304)

Customization

Metadata Configuration

Use a TOML file (e.g., metadata.toml) to provide information absent from GFF3/FASTA files, such as submitter details and common qualifiers.

• Example: See metadata_ddbj_example.toml.
• Default: If --metadata is omitted, the tool uses this default configuration.

Key Sections

1. COMMON Entry: Define SUBMITTER, REFERENCE, and COMMENT blocks.

2. Global Qualifiers (DDBJ-side injection): Use the [COMMON.feature] syntax to instruct the DDBJ system to insert qualifiers into every occurrence of a feature.

   [COMMON.assembly_gap]
   estimated_length = "unknown"
   gap_type = "within scaffold"
   linkage_evidence = "paired-ends"

3. Local Injection (Tool-side injection): Use the [feature] syntax (without the COMMON prefix) to have gff3-to-ddbj explicitly insert these qualifiers into the generated .ann file.

   [assembly_gap]
   estimated_length = "<COMPUTE>"  # Automatically calculate gap size from "N" runs
   gap_type = "within scaffold"
   linkage_evidence = "paired-ends"

   Note: Currently, only [source] and [assembly_gap] are supported for local injection.

[Advanced] Feature and Qualifier Renaming

GFF3 and DDBJ formats do not share a 1:1 nomenclature. GFF3 “types” (column 3) map to DDBJ “Features,” while GFF3 “attributes” (column 9) map to DDBJ “Qualifiers.”

gff3-to-ddbj uses a default translation table to handle these conversions. You can override these rules using --config_rename <FILE>.

Customization Examples:

• Renaming Types: Map a GFF3 type to a specific DDBJ feature key.

  [five_prime_UTR]
  feature_key = "5'UTR"

• Renaming Attributes: Map GFF3 attributes to DDBJ qualifiers. Use __ANY__ to apply a rule across all feature types.

  [__ANY__.ID]
  qualifier_key = "note"
  qualifier_value_prefix = "ID:"  # optional

• Complex Translations: Map a GFF3 type to a DDBJ feature/qualifier pair (e.g., snRNA to ncRNA with a class).

  [snRNA]
  feature_key = "ncRNA"
  qualifier_key = "ncRNA_class"
  qualifier_value = "snRNA"

• Attribute-to-Feature Mapping: Convert specific attribute values into distinct DDBJ features (e.g., RNA type with biotype=misc_RNA attribute becomes a misc_RNA feature).

  [RNA.biotype.misc_RNA]
  feature_key = "misc_RNA"

[Advanced] Feature and Qualifier Filtering

To comply with the DDBJ usage matrix, output is filtered by a default configuration. Only features and qualifiers explicitly allowed in this TOML file will appear in the final output.

To use a custom filter, provide a TOML file via --config_filter <FILE> using the following structure:

# Only these qualifiers will be kept for the CDS feature
CDS = ["EC_number", "inference", "locus_tag", "note", "product"]

Troubleshooting

Validate GFF3

It might be a good practice to validate your GFF3 files. GFF3 online validator is useful though the file size is limited to 50MB.

Split FASTA from GFF3 (if needed)

GFF3_to_DDBJ does not work when GFF3 contains FASTA information inside with ##FASTA directive. Attached tool split-fasta reads a GFF3 file and saves GFF3 (without FASTA info) and FASTA.

split-fasta path/to/myfile.gff3 --suffix "_splitted"

This creates two files, myfile_splitted.gff3 and myfile_splitted.fa.

Normalize entry names (if needed)

Letters like =|>" [] are not allowed in the 1st column (= “Entry”) of the DDBJ annotation. The attached program normalize-entry-names renames such entries. This program converts an ID like ERS324955|SC|contig000013 into ERS324955:SC:contig000013 for example.

normalize-entry-names myannotation_output.txt

This command create as files myannotation_output_renamed.txt if the invalid letters are found. Otherwise, you’ll see no output.

Known Issues

Biological & Sequence Logic

• Trans-splicing: The tool does not currently support coordinate correction or the join() syntax for features containing the /trans_splicing qualifier.
• Translation Exceptions: Coordinate handling for /transl_except at start or stop codons is not yet implemented.
• Missing Qualifiers: The tool does not automatically generate a /translation qualifier when an /exception qualifier is present, which may lead to DDBJ validation errors.
• Inter-base Coordinates: “Between-position” locations (e.g., 123^124) are not currently supported and may be processed incorrectly.

Performance

• Execution Speed: To ensure maximum accuracy, the tool currently utilizes a single-process architecture. Expect longer runtimes on large genomic datasets.

Acknowledgments

The design of GFF3-to-DDBJ is inspired by EMBLmyGFF3, a versatile tool used for converting GFF3 data into the EMBL annotation format.