v2.3.1 - Benchmarking suite

[v2.3.1] - 2026-04-12

Added

• New benchmarking directory, to test out the pipeline with bacterial, insect and mammalian input, and create various plots to demonstrate time and sucess metrics.
• Working example input datasets for the benchmarking

Changed

• Improved the ncbi genome download module to better handle issues in the input spreadsheet like trailing whitespace that was causing errors.
• Reduced the allowed forks of ncbi genome download to help prevent throttling.

v2.3.0 - GO fix variation

[v2.3.0] - 2026-04-08

Added

• GO enrichment results (CAFE_GO_RUN, CAFE_GO_PREP_LARGE, CAFE_GO_RUN_LARGE, CHROMO_GO, SUMMARIZE_CHROMO_GO) are now published to subfolders named after the active GO settings (e.g. cafe_go/weight01_t_cutoff0.05_typenone/). Re-running with different --go_algo, --go_cutoff, or --go_type values will write to a separate subfolder, preserving results from all parameter combinations.
• SUMMARIZE_CHROMO_GO now handles the case where no significant GO terms are found (no *_res.tab files produced by CHROMO_GO). The process exits cleanly with a message instead of erroring, and its outputs are marked optional so the pipeline continues.
• CAFE_MODEL_COMPARE now publishes the winning model's CAFE5 results directory to results/cafe/best/, mirroring the layout of results/cafe/base/. The complex channel derivation of ch_best_results in the workflow is replaced by a direct CAFE_MODEL_COMPARE.out.best_results emit.
• New OG_ANNOTATION_SUMMARY module: when --run_eggnog is set, produces OG_annotation_summary.tsv with one row per orthogroup containing the representative gene ID, description, preferred name, COG category, KEGG KO, and PFAM domains from EggNOG-mapper annotations. Output is written to results/eggnogmapper/OG_annotation_summary.tsv.
• New --eggnog_rep_species parameter to force the representative species used for orthogroup annotation. When unset, the species with the most annotated genes is chosen automatically.
• EGGNOGMAPPER output (.emapper.annotations files) is now published to results/eggnogmapper/.
• CAFE_RUN_LARGE now emits an optional converged.txt sentinel. CAFE_PLOT_LARGE and CAFE_GO_PREP_LARGE only run when convergence was achieved, preventing downstream failures when high-differential families cannot be modelled.
• CAFE_RUN_LARGE lambda retry sweep extended: now tries estimated λ, 0.005, 0.001, 0.0005, 0.0001, 0.00005, 0.00001, 0.000001, 0.0000001 (previously only descended; now also tries larger values which are more physically appropriate for families with extreme size differentials).
• EGGNOG_TO_OG_GO now parallelises GO file reading using ProcessPoolExecutor with task.cpus workers, reducing runtime on large species sets.

Changed

• EGGNOG_TO_OG_GO label changed from process_single to process_high (8 CPUs) to support parallel GO file reading.
• Pipeline version banner now reads dynamically from manifest.version in nextflow.config instead of being hardcoded in main.nf.
• cafe_max_k and eggnog_rep_species added to nextflow_schema.json (were defined in nextflow.config but missing from the schema, causing validation warnings).

Fixed

• OG_ANNOTATION_SUMMARY was incorrectly triggered when --run_eggnog was false (condition was run_eggnog || !skip_cafe), causing a "No such variable: ch_annot_files" error. The process now only runs when --run_eggnog is set.

v2.2.0 - Easter Rising

[v2.2.0] - 2026-04-06

Added

• New OrthoFinder algorithm parameters: --orthofinder_method (-M), --orthofinder_search (-S), --orthofinder_msa_prog (-A), and --orthofinder_tree (-T). These map directly to OrthoFinder command-line flags and are all optional — OrthoFinder defaults are used when unset.
• New --orthofinder_v2 flag (default false) to run OrthoFinder v2.5.5 instead of v3.1.3. v2 uses Hierarchical Orthogroups (N0.tsv) which are more appropriate for CAFE5 as they represent gene families traceable to the common ancestor. v3 uses flat orthogroups (Orthogroups.tsv) which can have inflated copy-number variance. For large datasets (>30 species), v2 is recommended.
• ORTHOFINDER_V2_CAFE results are now published to results/orthofinder_cafe/ (was only published for v3).
• CAFE_PREP now emits pruned_tree (the rescaled, species-name-stripped tree) for use by all downstream CAFE runs.
• CAFE_RUN_LARGE now retries with progressively smaller lambda values (estimated → 1e-4 → 1e-5 → 1e-6 → 1e-7) when the initial fixed-lambda run fails to converge, as recommended in hahnlab/CAFE5#132.
• CAFE GO enrichment plots now display full GO term text labels and GO IDs.
• New output documentation page docs/outputs.md with example figures and detailed descriptions of all output files.

Changed

• Reverted tree scaling back to the original RESCALE_TREE approach (rescale_tree.py multiplies branch lengths by --tree_scale_factor). The MAKE_ULTRAMETRIC module introduced in v2.1.4 is removed.
• --tree_scale_factor default changed from 1 back to 1000.
• CAFE_PREP base run and error model estimation now use pruned_tree (the rescaled non-ultrametric tree) directly, matching the approach that was validated on large datasets. SpeciesTree_rooted_ultra.txt (produced by chronoMPL()) is retained for reference only.
• CAFE_RUN_K and CAFE_RUN_BEST now use pruned_tree instead of SpeciesTree_rooted_ultra.txt, avoiding convergence failures caused by the ultrametric tree's maximum-possible-lambda constraint.
• CAFE_RUN_LARGE failure is now non-fatal — the pipeline continues even if high-differential families cannot be modelled.
• CAFE_PLOT (and CAFE_PLOT_LARGE) now skip gracefully when CAFE5 did not produce an *_asr.tre file, instead of crashing the pipeline.
• ORTHOFINDER_V2 module now emits N0.tsv as orthologues (previously emitted Orthogroups.tsv). This ensures CAFE_PREP receives hierarchical orthogroups, which have lower copy-number variance and are required for correct CAFE5 analysis.
• CHROMO_GO now symlinks N0.tsv to Orthogroups.tsv when the v2 path is used, so the downstream perl script works regardless of OrthoFinder version.

Fixed

• Fixed species name mismatch between tree and gene counts in CAFE_RUN_K: the SpeciesTree_rescaled.nwk retains .clean suffixes but hog_gene_counts.tsv uses plain names. All CAFE runs now use pruned_tree which has suffixes stripped by sed in CAFE_PREP.
• Fixed chronos() convergence failure on large datasets: passing the ×1000 pre-scaled tree to chronos() caused a degenerate starting point (~-74 billion log-likelihood). CAFE_PREP now receives the original unscaled tree and scaling is applied after any ultrametric correction.

v2.1.1 Augustus corrected

[v2.1.1] - 2026-04-01
Fixed
Allow Augustus transcript IDs to output to GO hash
Force eggnogmapper to process_high only, High memory causes stalling issues

v2.1.0 Fully parallel

[v2.1.0] - 2026-03-30

New features

• CAFE runs in three separate jobs
• CAFE tests which run was the most likely. []. GO running on that.
• CAFE GO run in parrelel.
• Nextlfow schema added (replacing nf-validation)

Changed

• New readme output file format added.

Fixed

• Fixed issue with flag names for eggnog parameters.

v2.0.4 Minor patch

Had to fix bug in conf/modules.config

v2.0.3 - Module Upgrade

What's Changed

• Transcript g os by @chriswyatt1 in #71

Full Changelog: v2.0.2...v2.0.3

v2.0.2 - Parellel and GO filter

Add new filter flag options

What's Changed

• Chromo go parellelise by @chriswyatt1 in #68
• Add option flags for eggnogmapper, main ones by @chriswyatt1 in #69

Full Changelog: v2.0.1...v2.0.2

v2.0.1 - Augustus adapted

Fixed major issues with alternative gff types.

Added function to adapt names of samples from AUGUSTUS type GFFs.
Fixed missing GO plot errors.
Schema and config updates to new modules and usage
New gffmix input test to make sure it works with a variety of gff types.
Added missing containers for some modules, and with ps to work with nextflow

v2.0.0 - GO_improve

[v2.0.0] - 2026-03-20

New features

• Added EggNOG-mapper integration for GO annotation (--run_eggnog)
• Added automatic EggNOG database download (EGGNOG_DOWNLOAD)

Enhancements

• Updated OrthoFinder module to v3.1.3 with corrected output paths
• Added AGAT_CONVERTSPGXF2GXF step for robust GFF standardisation
• Improved CAFE R scripts to support both OrthoFinder v2 and v3 output formats
• Simplified workflow — removed unused modules (GET_DATA, GO_ASSIGN, GO_EXPANSION)
• CHROMO_GO now uses ORTHOFINDER_CAFE results instead of waiting for ORTHOFINDER_GO

Bug fixes

• Fixed meta map handling for NCBIGENOMEDOWNLOAD outputs
• Fixed stop codon (.) removal from protein fastas before OrthoFinder/EggNOG
• Fixed EggNOG database download. Using wget instead, which is stable.
• Fixed OrthoFinder v3 output path changes (Orthogroups/ vs WorkingDirectory/Orthogroups/)
• Fixed cafe_go.pl and cafe_prep.R for OrthoFinder v3 column name changes

Parameters added

• --run_eggnog — enable GO annotation via EggNOG-mapper
• --eggnog_data_dir — path to pre-downloaded EggNOG database