← Về thư mục
📄 / / profiles / r-and-d / raw_evidence / OpenAI4S / skills / bioskills / bio-genome-annotation-prokaryotic-annotation / references / usage-guide.md

Prokaryotic Annotation - Usage Guide

Overview

Annotate bacterial and archaeal genomes (isolates, MAGs, plasmids) with structural and functional gene predictions using Bakta (default) or Prokka (legacy), with NCBI PGAP and DFAST as the submission-grade paths. Produces GFF3, GenBank/EMBL, protein FASTA, and a feature table ready for downstream analysis or INSDC submission. The hard part is not finding genes (callers recover >95%) but assigning trustworthy function, picking the right translation table, and not comparing gene counts across mismatched tools or database versions.

Prerequisites

# Bakta (default) + database (record the version)
conda install -c bioconda bakta
bakta_db download --output /path/to/bakta_db --type full   # ~30 GB; --type light for triage/CI

# Prokka (legacy)
conda install -c bioconda prokka

# QC gate (run before trusting annotation numbers)
conda install -c bioconda checkm2

# Python parsing
pip install gffutils biopython

Quick Start

Tell your AI agent what you want to do: - "Annotate my bacterial genome assembly with Bakta and report coding density" - "Annotate my Mycoplasma genome with the correct translation table" - "Re-annotate this collection of Klebsiella genomes uniformly for pangenomics" - "Prepare my genome annotation for NCBI submission"

Example Prompts

Basic Annotation

"Annotate this bacterial assembly with Bakta using the full database, gram-negative, locus tag prefix SAUR, and give me the GFF3 plus a coding-density and hypothetical-protein QC."

"Run Bakta on my Staphylococcus genome and flag whether the coding density and gene count are in the expected range."

Organism-Specific Cases

"Annotate my Mycoplasma genitalium assembly - it uses genetic code 4 (UGA=Trp), so make sure the translation table is set correctly."

"Annotate this archaeon with Bakta and confirm the archaeal tRNA model was used."

Collections and Submission

"Re-annotate all of these E. coli assemblies from FASTA with one Bakta version and DB so the pangenome accessory genome isn't inflated by annotation drift."

"Prepare my Bakta annotation for NCBI GenBank submission with INSDC-compliant locus tags."

What the Agent Will Do

  1. Gate on assembly quality first (CheckM2 completeness/contamination); flag if contamination >5% or completeness <90%
  2. Determine the correct translation table from taxonomy (GTDB-Tk if available), not a guess
  3. Run Bakta (or Prokka/PGAP/DFAST per the target) with organism-specific parameters and a recorded DB version
  4. Generate GFF3, GenBank, protein FASTA, and summary outputs
  5. Report annotation statistics (gene count, coding density, tRNA/rRNA counts, hypothetical fraction)
  6. Flag failure signatures (low coding density -> wrong table/fragmentation; pseudogene spike -> homopolymer/phase-variation/decay; near-0% hypothetical -> over-confident transfer)
  7. Suggest functional-annotation follow-up (eggNOG-mapper/InterProScan) for hypotheticals

Tips