CDScompare

CDScompare performs the quantitative comparison of structural genome annotations (GFF3) of the same genome.
For each pair of overlapping genes, it computes a frame-aware CDS similarity score based on coding-region agreement and reading-frame conservation.
It is implemented in Python and distributed as a command-line tool.

The tool supports:

• pairwise comparison between two genome annotations
• multi-comparison of several annotations against a common reference
• two pairing strategies (best or all)

Installation

Requirements

One of the following must be available:

• Python ≥ 3.10 with pip or pipx
• Docker or Apptainer / Singularity

---

Installation from PyPI (recommended)

Using pipx (isolated CLI install)

pipx install cdscompare==<version>

Replace <version> with a specific release number (e.g. 0.3.0rc5).

Using pip

pip install cdscompare==<version>

Test the installation:

cdscompare --version

Container image (Docker / Apptainer)

A pre-built OCI image is available via GitHub Container Registry:

ghcr.io/johgi/cdscompare:<version>

Docker

Pull the image:

docker pull ghcr.io/johgi/cdscompare:<version>

To run on local files:

docker run --rm \
  -v "$PWD:/work" \
  -w /work \
  ghcr.io/johgi/cdscompare:<version> \
  file1.gff file2.gff

Apptainer / Singularity

Pull the image:

apptainer pull docker://ghcr.io/johgi/cdscompare:<version>

To run on local files:

apptainer run \
  --bind "$PWD:/work" \
  cdscompare_<version>.sif \
  /work/file1.gff /work/file2.gff

Installing development / unreleased versions

git clone git@github.com:ranwez/CDScompare.git
cd CDScompare
pip install .

After installation, the repository is no longer required.

---

Preparing input GFF files

CDScompare expects both input annotation files to describe protein-coding gene models in a consistent GFF3-like structure. Because GFF/GTF files may differ substantially between annotation tools and databases, we strongly recommend standardizing input files before running CDScompare.

In particular, CDScompare assumes that all input annotation files:

• describe the same genome assembly;
• follow standard GFF3 conventions, including:
  • tab-delimited records with 9 columns;
  • gene, mRNA and CDS features organized in a consistent gene -> mRNA -> CDS hierarchy;
  • valid ID and Parent attributes linking genes, transcripts and CDS features;
  • unique gene and mRNA identifiers within each file;
  • coherent genomic coordinates;
  • non-overlapping CDS features within each transcript;
  • valid CDS phases (0, 1, 2, or .);
• use compatible seqids in column 1 across annotations.

Seqid compatibility

The seqids in column 1 of the two GFF files must be compatible. CDScompare can harmonize an optional leading chr prefix when this improves the number of shared seqids and does not merge distinct seqids within the same annotation. For example, annotations using chr1, chr2, chr3 can therefore be compared to annotations using 1, 2, 3.

More complex naming differences, such as NC_000001.11 vs chr1, chromosome_1 vs 1, or scaffold-specific naming schemes, must be harmonized by the user before running CDScompare.

Recommended preprocessing with AGAT

Before running CDScompare, we strongly recommend standardizing each input annotation with AGAT (Dainat, 2022). For example:

agat_convert_sp_gxf2gxf.pl -g annotation.gff3 -o annotation.agat.gff3

AGAT can help standardize GFF/GTF files, sort features, fix duplicated IDs, add missing ID and Parent attributes, add missing parent features when possible, and produce a more consistent GFF3 structure. This preprocessing step is especially recommended when input files come from different annotation tools or databases.

Scope and assumptions

• Gene overlap is detected based on gene genomic coordinates.
• Similarity scores are computed only from CDS features.
• Only gene models represented as a gene -> mRNA -> CDS hierarchy are considered. Genes without CDS-containing mRNAs, transcripts without CDS features, and features not represented as gene entries are ignored.
• Alternative splicing is handled by selecting the best-matching mRNA pair for each gene pairing.

---

Command-line interface

cdscompare [OPTIONS] ANNOT1_GFF ANNOT2_GFF [ANNOT3_GFF ...]

Positional arguments

Argument	Description
ANNOT1_GFF	First annotation file, reported as annotation 1 in pairwise output
ANNOT2_GFF	Second annotation file, reported as annotation 2 in pairwise output
ANNOT3_GFF ...	Optional additional annotation files, each compared independently against ANNOT1_GFF

Note:

• At least two GFF files are required.
• In pairwise comparisons (exactly two GFF files), similarity scores and gene pairings are invariant to the input file order.
• GFF input file basenames must be unique (used as annotation identifiers).

Options

Option	Description
-d, --out_dir	Output directory where result files are written (default: results).
-p, --pairing_mode	Pairing strategy used within clusters of overlapping genes. Possible values are: • best (default): selects a globally optimal gene pairing using dynamic programming. • all: reports all overlapping gene pairings without global optimization.

Output files

Pairwise comparison

For each pairwise comparison, CDScompare produces two output files.

1. Detailed comparison file (<annotation1>_vs_<annotation2>.csv)

This file contains one row per reported gene pair or unpaired gene, with the following columns:

Column	Description
seqid_strand	Input GFF seqid followed by the _direct or _reverse strand suffix
cluster	Identifier of the cluster of overlapping genes
annot1_gene	Gene identifier in annotation 1
annot2_gene	Gene identifier in annotation 2
matches	Number of nucleotide positions that are coding in both annotations and in the same reading frame
mismatches	Total number of mismatched nucleotides (coding/non-coding + reading frame mismatches)
similarity_score	Frame-aware CDS similarity score (%), computed as 100 × matches / (matches + mismatches)
annot1_start	Start coordinate of the annotation 1 gene
annot1_end	End coordinate of the annotation 1 gene
annot2_start	Start coordinate of the annotation 2 gene
annot2_end	End coordinate of the annotation 2 gene
annot1_mRNA	Identifier of the selected mRNA from annotation 1 used for the comparison
annot2_mRNA	Identifier of the selected mRNA from annotation 2 used for the comparison
C_NC_mismatch_zones	Genomic intervals where nucleotides are annotated as CDS in only one of the two annotations
RF_mismatch_zones	Genomic intervals where reading frames differ between annotations
C_NC_mismatches	Total length (in nucleotides) of coding/non-coding mismatches
RF_mismatches	Total length (in nucleotides) of reading frame mismatches
annot1_mRNA_count	Number of CDS-containing mRNAs for the corresponding gene in annotation 1
annot2_mRNA_count	Number of CDS-containing mRNAs for the corresponding gene in annotation 2

Special values:

• _ : undefined or not applicable
• ~ : no corresponding gene reported for this comparison. This occurs when:
  • no overlapping gene exists in the other annotation, or
  • the gene was paired with a different gene in best pairing mode.

2. Summary file (<annotation1>_vs_<annotation2>.txt)

This file contains overall counts for the corresponding pairwise comparison:

Comparison summary for <annotation1> vs <annotation2>:
- reported gene pairs: X
- unpaired genes in annotation 1: Y
- unpaired genes in annotation 2: Z

Multi-comparison mode

When more than two annotation files are provided, the first annotation is used as a common reference and compared independently with each subsequent annotation.

For each comparison, CDScompare generates the same pairwise CSV and summary files described above. In these files:

• annotation 1 refers to the first input annotation;
• annotation 2 refers to the annotation currently compared with it.

An additional global summary file (synthesis_<annotation1>.csv) is generated, with one line per gene from the reference annotation, and two columns for each compared annotation:

Column	Description
<reference>_gene	Gene identifier in annotation 1
<annotation>_gene	Best-matching gene from the compared annotation
<annotation>_similarity_score	Similarity score (%) for the corresponding gene pairing

Citation

CDScompare has been accepted for publication in Plant Methods. Full citation details will be added when available.

Developers notes

Development installation

This method installs the package in editable mode and includes development dependencies (pytest and ruff).

git clone git@github.com:ranwez/CDScompare.git
cd CDScompare
pip install -e .[dev]

To run the tests:

python -m pytest -v

Developer documentation

The internal code structure documentation is generated using Sphinx.

To build the documentation locally:

pip install -e .[docs]
python -m sphinx -b html docs/source docs/build/html

The generated documentation can be opened locally in a web browser (docs/build/html/index.html).