Gene Family Evolution - Usage Guide

Overview

Gene-family birth-death models on phylogenies (Hahn 2005; Csurös 2010) detect lineage-specific gene-family expansions and contractions. CAFE5 (Mendes 2020 Bioinformatics 36:5516) is the modern standard with gamma-distributed rate categories for biologically realistic modeling. Annotation heterogeneity is the dominant confounder -- different annotation pipelines predict different gene counts, producing apparent "expansions" that are pipeline artifacts. Consistent annotation + BUSCO/Compleasm completeness filtering are mandatory.

For HGT-affected gene families (prokaryotes), CAFE5's birth-death-only model misses transfer; use ALE/GeneRax/AleRax instead ([[gene-tree-species-tree-reconciliation]]).

Prerequisites

# Modern CAFE5
conda install -c bioconda cafe

# Alternative methods
# Count (Csurös 2010; Java)
wget http://www.iro.umontreal.ca/~csuros/gene_content/count.tar.gz && tar xf count.tar.gz

# BadiRate (Librado 2012)
git clone https://github.com/PauloRoldan/badirate

# Tree time-calibration
conda install -c bioconda treepl  # or ape::chronos in R
# Or LSD2 (faster)

# Time-source database
# TimeTree (web): http://www.timetree.org

# Functional enrichment
Rscript -e "
install.packages('BiocManager')
BiocManager::install(c('clusterProfiler', 'org.Hs.eg.db', 'topGO'))
"

# OrthoFinder input
conda install -c bioconda orthofinder

Quick Start

Tell the AI agent what to do: - "Run CAFE5 on OrthoFinder HOG counts to identify gene families with lineage-specific expansions" - "Identify ancestral gene-family counts at internal nodes using Count" - "Run CAFE5-error mode with explicit annotation error rates per species" - "Test for trait-correlated rate shifts in gene families using RERconverge"

Example Prompts

Standard CAFE5 Analysis

"I have OrthoFinder HOG counts for 30 mammalian species. Build a time-calibrated species tree with treePL using divergence dates from TimeTree. Run CAFE5 with 4 gamma rate categories and identify families with lineage-specific rate shifts (p < 0.05, FDR-corrected). Annotate top expanded / contracted families with functional categories."

Annotation-Heterogeneity Correction

"The 30 mammalian genomes have heterogeneous annotation pipelines (some annotated with NCBI RefSeq, others with Ensembl, others with BRAKER3). Run CAFE5-error mode with empirical per-species annotation error rates. Compare results with strict CAFE5 to assess heterogeneity impact."

Convergent Rate Shifts

"Identify gene families with rate shifts in echolocating species (bats and dolphins) using RERconverge on the CAFE5 family rates. Cross-validate with CSUBST for convergent substitution patterns. Report top convergent gene families with functional annotation."

What the Agent Will Do

1. Validate inputs: OrthoFinder HOG matrix; species tree; per-genome BUSCO completeness
2. Filter genomes with BUSCO < 90% (or document)
3. Normalize annotation pipeline if possible; report error rates if not
4. Time-calibrate species tree with treePL / LSD2 / TimeTree
5. Run CAFE5 with gamma rate categories (-k 4)
6. CAFE5-error mode if annotation heterogeneity is present
7. FDR-correct across families
8. Identify lineage-specific expansions / contractions at significant rate shifts
9. Functional enrichment with clusterProfiler / topGO
10. Cross-validate with Count for ancestral state reconstruction; ALE for HGT-affected families
11. Report: per-family lambda, significant families, functional categories, ancestral state estimates
12. Caveats: annotation heterogeneity, assembly fragmentation, HGT, polyploidy

Tips

• CAFE5 (Mendes 2020) is the modern standard; CAFE4 deprecated
• Filter input to multi-copy orthogroups (max count >= 2); single-copy contribute no information
• Tree MUST be ultrametric (time-calibrated); use treePL / LSD2 / ape::chronos
• Negative branch lengths cause failures; calibrate properly
• Require >= 100 families for CAFE5 (preferably > 1000)
• Annotation heterogeneity is the single largest confounder; re-annotate consistently if possible
• BUSCO completeness >= 90% mandatory; exclude or correct fragmented genomes
• Multi-testing correction: FDR (Benjamini-Hochberg) across families
• For HGT-affected bacterial families, switch to ALE / GeneRax / AleRax
• For polyploid species, assign subgenomes first; analyze each subgenome separately
• For convergent rate shifts, RERconverge complements CAFE5; CSUBST for amino-acid convergence
• Whale.jl native WGD modeling for WGD-affected lineages
• Report robust lambda with and without outlier families; manual annotation of outliers
• Time calibration from TimeTree is convenient; or compute from molecular clock

Related Skills

comparative-genomics/ortholog-inference - OrthoFinder HOG matrix is CAFE5 input comparative-genomics/gene-tree-species-tree-reconciliation - ALE per-family DTL, complement comparative-genomics/whole-genome-duplication - Post-WGD retention bias comparative-genomics/positive-selection - Selection within expanded families comparative-genomics/ancestral-reconstruction - Ancestral count reconstruction phylogenetics/divergence-dating - Time-calibrated tree for CAFE5 pathway-analysis/go-enrichment - Functional enrichment of expanded families pathway-analysis/gsea - GSEA on family expansions