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.