Hgt Detection

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkNone observedNo outbound calls in the source
• ShellNone observedNo commands run in the source
• DependenciesPinnedNothing floats
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass.

Audit 0.4.1 · source sha c8c1fc3a2b13 · 19 Oct 2026Report an issue or request a re-scan

• PublisherGPTomics GitHub organizationOwner of the repository this skill is read from
• Source folderGPTomics/bioSkills/comparative-genomics/hgt-detectionThe exact folder we audited
• LicenceMITAs the repository states it
• Commit pinned for this verdictc8c1fc3a2b13A new audit runs when the source changes
• Repository statusArchivedLast push 65 days ago
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: Detect horizontal gene transfer (HGT / LGT) using compositional methods (GC%, codon usage, tetranucleotide z-scores via SIGI-HMM, AlienHunter, IslandViewer 4, IslandPath-DIMOB), phylogenetic-incongruence methods (AvP,…

---
name: bio-comparative-genomics-hgt-detection
description: Detect horizontal gene transfer (HGT / LGT) using compositional methods (GC%, codon usage, tetranucleotide z-scores via SIGI-HMM, AlienHunter, IslandViewer 4, IslandPath-DIMOB), phylogenetic-incongruence methods (AvP, HGTphyloDetect, ALE / GeneRax / AleRax reconciliation, RANGER-DTL), and BLAST-distribution methods (HGTector v2, DarkHorse, Alien Index). Use when screening prokaryote genomes for genomic islands and HGT events, distinguishing HGT from incomplete lineage sorting / differential gene loss / hybridization, mapping donor lineages via phylogenetic placement, separating eukaryotic HGT from contamination, ruling out gBGC as a false signal, or quantifying DTL rates with ALE/GeneRax on bacterial trees.
tool_type: mixed
primary_tool: HGTector
---

## Version Compatibility

Reference examples tested with: HGTector 2.0b3+, AvP 1.0.4+, HGTphyloDetect 1.0+, ALE 1.0+ (ssolo/ALE github), GeneRax 2.1.3+, AleRax 1.2.0+ (Morel 2024), RANGER-DTL 2.0+, IslandViewer 4 (web), mobileOG-db 1.0+, MetaCHIP 1.10+, IQ-TREE 2.3.6+, BioPython 1.84+, DIAMOND 2.1.10+. Open Tree of Life and NCBI Taxonomy reference databases updated 2024-Q3 minimum for HGTector/AvP.

Before using code patterns, verify installed versions match. If versions differ:
- Python: `pip show hgtector` then `hgtector search --help`
- CLI: `ALEml_undated --help`, `generax --help`, `alerax --help`
- DB: `hgtector database --check` for taxonomy version

If code throws `Taxonomy ID not found`, `database version mismatch`, or `KeyError` on NCBI taxids, refresh the local taxonomy dump (NCBI updates monthly). ALE/GeneRax expect newick gene trees with bootstraps; AleRax expects gene-tree distributions (uniform bootstrap samples or UFBoot trees).

# Horizontal Gene Transfer Detection

**"Are these genes horizontally acquired, and from where?"** -> HGT signal lives in three orthogonal signal classes: composition (recent transfers carry donor codon usage; erodes by Lawrence-Ochman 1998 amelioration in ~50-200 Myr), phylogeny (gene tree nests within distant clade), and phyletic distribution (patchy taxonomic presence). No single class proves HGT; **claims require concordance across at least two classes** plus mandatory exclusion of contamination and differential gene loss (DGL). The most consequential failure mode in eukaryotic HGT detection is contamination passing all three classes silently (Boothby 2015 tardigrade "17% HGT" refuted by Koutsovoulos 2016; Crisp 2015 human "145 HGTs" refuted by Salzberg 2017 GB 18:85).

- Python: `hgtector search` -> `hgtector analyze` for BLAST-distribution screen
- Python: `AvP` (Koutsovoulos 2022 PLoS Comp Biol 18:e1010686) for eukaryotic phylogenetic HGT with automated tree workflow
- CLI: `ALEml_undated` (Szöllősi 2013 Syst Biol 62:901), `generax` (Morel 2020 MBE 37:2763), `alerax` (Morel 2024 Bioinformatics 40:btae162) for prokaryote DTL reconciliation
- Web: IslandViewer 4 (Bertelli 2017 NAR 45:W30) for bacterial genomic islands
- CLI: `metachip` (Song 2019 Microbiome 7:36) for metagenomic HGT inference

## Algorithmic Taxonomy

| Method class | Tool | Signal | Strength | Fails when |
|--------------|------|--------|----------|------------|
| Composition (parametric) | SIGI-HMM (Waack 2006 BMC Bioinf 7:142) | HMM on codon-usage anomaly | Recent transfers (<50 Myr); single-locus resolution | Amelioration eroded composition; native composition heterogeneous (Streptomyces, Borrelia) |
| Composition (parametric) | AlienHunter (Vernikos & Parkhill 2006 Bioinformatics 22:2196) | Variable-window tetranucleotide IVOM | Recent island detection; window-size aware | Old transfers; high-variance native composition |
| Composition (parametric) | IslandPath-DIMOB (Bertelli 2017) | Dinucleotide bias + mobility genes | Combines composition + mobile-element signature | Mobile-element-free transfers |
| Composition aggregator | IslandViewer 4 web (Bertelli 2017 NAR 45:W30) | Consensus of IslandPath / SIGI / IslandPick | Best single-genome bacterial screen; curated benchmark | Recent radiations; close-relative donors |
| BLAST-distribution | HGTector v2 (Zhu 2014 BMC Genomics 15:717) | Close vs distal BLAST hit ratio against full taxonomy | No tree required; scales to thousands of genomes | Close-relative donors (similar lineage hits); incomplete taxonomic sampling |
| BLAST-distribution | DarkHorse (Podell & Gaasterland 2007 GB 8:R16) | Lineage Probability Index (LPI) | Quantifies taxonomic prior of best hits | Same lineage-coverage caveat as HGTector |
| BLAST-distribution | Alien Index (Gladyshev 2008 Science 320:1210; recent: AI tools) | Best hit metazoan vs non-metazoan score | Standard for eukaryote HGT screen | False positives from rapid evolution / contamination |
| Phylogenetic incongruence | AvP (Koutsovoulos 2022) | Automated tree-building + AU test on candidate HGTs | End-to-end eukaryote pipeline; orthogroup-aware | Poor taxon sampling at putative donor lineage |
| Phylogenetic incongruence | HGTphyloDetect (Yuan 2023 Brief Bioinform 24:bbad035) | Web-friendly; Bayesian incongruence test | Modern eukaryote-friendly | Computationally heavy at genome scale |
| Probabilistic DTL reconciliation | ALE (Szöllősi 2013) | Amalgamated likelihood over gene-tree sample | Bayesian-posterior over D/T/L events; explicit donor inference | Requires gene tree distribution (bootstrap or UFBoot) |
| Probabilistic DTL reconciliation | GeneRax (Morel 2020) | ML reconciliation; species-tree-aware | Faster than ALE; refines noisy gene trees | Less uncertainty quantification than ALE |
| Probabilistic DTL reconciliation | AleRax (Morel 2024) | Co-estimates gene + species trees + DTL rates | Gold standard 2024; corrects gene-tree-error feedback | Computationally heavy; needs >= 20 species |
| Parsimony reconciliation | RANGER-DTL 2.0 (Bansal 2018 Bioinformatics 34:3214) | Min-cost DTL parsimony | Fast; deterministic; many trees | No likelihood; sensitive to event-cost choice 

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

pip install hgtector
git clone https://github.com/GDKO/AvP && cd AvP && pip install .
git clone --recursive https://github.com/BenoitMorel/AleRax && cd AleRax && ./install.sh
pip install conterminator
git clone https://github.com/clb21565/mobileOG-db

git clone https://github.com/GPTomics/bioSkills
git -C bioSkills checkout c8c1fc3a2b13

Works in

1. 2026-10-19
   ✓SAFE B 89 / 100

   first audit · source c8c1fc3a2b13

Questions about Hgt Detection

• github.com/GPTomics/bioSkills/tree/main/comparative-genomics/hgt — the audited source at commit c8c1fc3a2b13 and every finding aboveObserved · 19 Oct 2026

Spotted something wrong? Report a correction · Methodology