Ortholog Inference - Usage Guide

Overview

Ortholog inference distinguishes genes whose history reflects speciation (orthologs) from genes derived from duplication (paralogs). HOGs (Hierarchical Orthologous Groups; Altenhoff 2013) are the modern unit of analysis -- a maximal cluster of genes descended from a single ancestral gene at a defined taxonomic level. OrthoFinder3 (Emms et al 2026 Nat Methods) is the current Quest-for-Orthologs benchmark leader; its HOG output is ~7% more accurate than v2 orthogroups (5-7% higher f-score); outgroup inclusion further improves accuracy. For prokaryote-eukaryote scale, TOGA (Kirilenko 2023 Science) integrates whole-genome-alignment-based orthology with gene-loss classification.

The "ortholog conjecture" (orthologs more functionally similar than paralogs) is supported but weakly (Altenhoff 2012), though Stamboulian 2020 finds paralogs are often comparable predictors of function. Don't treat 1:1 ortholog labeling as automatic functional equivalence.

Prerequisites

# Modern tools
conda install -c bioconda orthofinder=3.0
conda install -c bioconda sonicparanoid broccoli proteinortho

# OMA / FastOMA
wget https://omabrowser.org/standalone/OMA.tgz && tar xf OMA.tgz
pip install fastoma

# eggNOG-mapper
conda install -c bioconda eggnog-mapper

# JustOrthologs (close species, fast)
git clone https://github.com/ridgelab/justOrthologs

# TOGA (WGA-anchored orthology + gene loss)
conda env create -f https://raw.githubusercontent.com/hillerlab/TOGA/master/toga_env.yml

# Quality control
conda install -c bioconda busco compleasm

Quick Start

Tell the AI agent what to do: - "Run OrthoFinder3 on these proteomes with outgroup for HOG inference" - "Cross-validate with SonicParanoid2 for consensus single-copy orthologs" - "Identify ortholog vs paralog (in-paralog vs co-ortholog) for gene X across vertebrates" - "Project annotations from human to chimp using TOGA with intactness classification"

Example Prompts

Standard HOG Inference

"Run OrthoFinder3 on these 30 proteomes (vertebrates with outgroup invertebrate). Pre-filter to longest isoform per gene. Extract single-copy HOGs for phylogenomic concatenation. Cross-validate with SonicParanoid2 and report consensus single-copy set."

Functional Annotation Transfer

"Transfer functional annotations from human Ensembl release 113 to a newly sequenced primate genome. Use eggNOG-mapper for primary annotation + OrthoFinder3 HOGs for cross-validation. Apply ortholog-conjecture-aware confidence stratification: 1:1 ortholog + low dN/dS + EXP evidence = high confidence."

Vertebrate-Scale Orthology with Gene Loss

"Identify orthology and gene-loss across 100 mammal genomes using TOGA + CESAR. First build Cactus WGA. Then TOGA Nextflow pipeline. Report per-genome intactness distribution (I/PI/UL/L/M/PM). Identify mammalian lineages with notable gene loss."

What the Agent Will Do

1. Validate inputs: BUSCO/Compleasm completeness per species; consistent annotation pipeline
2. Pre-filter to longest isoform per gene (AGAT or OrthoFinder primary_transcript.py)
3. Include outgroup at next-higher taxonomic level (essential for HOG rooting)
4. Run primary tool: OrthoFinder3 default for most cases; SonicParanoid2 for cross-validation
5. Extract HOG at appropriate level: N0 root for phylogenomic; per-clade levels for clade-specific
6. Filter single-copy for concatenation: present in >= 90% species, 1 copy each
7. Classify orthology: 1:1, 1:many, many:many; in-paralog vs out-paralog
8. For functional transfer: eggNOG-mapper + 1:1 + dN/dS < 0.2 + EXP evidence
9. For WGA-anchored (vertebrate-scale): TOGA + CESAR
10. Report: per-species classification, orthogroup statistics, consensus single-copy set
11. Caveats: WGD effects, ortholog conjecture, annotation heterogeneity, hidden paralogy

Tips

• OrthoFinder3 (2026 Nat Methods) is the modern Quest-for-Orthologs leader
• HOG output (v3) is at Phylogenetic_Hierarchical_Orthogroups/N0.tsv; v2 layout differs
• Include >= 1 outgroup at next-higher taxonomic level for STRIDE rooting (improves HOG accuracy)
• Pre-filter to longest isoform per gene; otherwise isoforms inflate co-ortholog counts
• For WGD-affected lineages, use synteny-aware methods (GENESPACE, ProteinOrtho-synteny)
• Cross-validate with 2+ methods (OrthoFinder3 + SonicParanoid2 or OrthoFinder3 + OMA)
• BUSCO/Compleasm completeness >= 90% required for inclusion
• TOGA (Kirilenko 2023) for vertebrate-scale orthology with explicit gene-loss classification
• Annotation pipeline normalization across species; mixing pipelines inflates accessory
• Ortholog conjecture is weakly supported; treat 1:1 ortholog functional transfer with care
• For HGT-affected lineages, OrthoFinder may return "vertical orthologs" that aren't real; use ALE
• eggNOG-mapper integrates orthology + functional annotation; current eggNOG database mandatory
• JustOrthologs is fastest for closely related species (same family); OrthoFinder3 for broader

Related Skills

comparative-genomics/synteny-analysis - Synteny-anchored disambiguation in WGD lineages comparative-genomics/whole-genome-duplication - Distinguishing ohnologs from orthologs comparative-genomics/gene-family-evolution - CAFE5 birth-death on OG counts comparative-genomics/gene-tree-species-tree-reconciliation - DTL-aware orthology for prokaryotes comparative-genomics/positive-selection - Selection on single-copy alignments comparative-genomics/comparative-annotation-projection - TOGA / CESAR uses orthology phylogenetics/modern-tree-inference - Single-copy concatenation phylogenomics phylogenetics/species-trees - Coalescent species tree from gene trees alignment/multiple-alignment - MSA quality affects HOG output genome-annotation/functional-annotation - eggNOG / orthology-based annotation read-qc/quality-reports - BUSCO / Compleasm completeness