Version Compatibility

Bakta

Database Setup

# Download the full database (~30 GB, recommended for comprehensive annotation)
bakta_db download --output /path/to/bakta_db --type full

# Lightweight database (~1.5 GB, faster but less comprehensive)
bakta_db download --output /path/to/bakta_db --type light

# Update existing database
bakta_db update --db /path/to/bakta_db

Basic Annotation

bakta \
    --db /path/to/bakta_db \
    --output bakta_out \
    --prefix my_genome \
    --locus-tag MYORG \
    --threads 8 \
    assembly.fasta

Key Options

With Organism Metadata

bakta \
    --db /path/to/bakta_db \
    --output bakta_out \
    --prefix ecoli_k12 \
    --locus-tag ECK12 \
    --genus Escherichia --species coli --strain K-12 \
    --gram - \
    --complete \
    --threads 16 \
    assembly.fasta

Output Files

bakta_out/
├── my_genome.gff3       # GFF3 annotation (primary output)
├── my_genome.gbff       # GenBank format
├── my_genome.ffn        # Nucleotide CDS sequences
├── my_genome.faa        # Protein sequences
├── my_genome.fna        # Annotated genome sequence
├── my_genome.embl       # EMBL format
├── my_genome.tsv        # Tab-separated feature table
├── my_genome.json       # Machine-readable JSON
└── my_genome.txt        # Summary statistics

Prokka (Legacy Alternative)

Prokka is lighter weight and faster but uses older databases. Prefer Bakta for new projects.

prokka \
    --outdir prokka_out \
    --prefix my_genome \
    --locustag MYORG \
    --genus Escherichia --species coli \
    --cpus 8 \
    --rfam \
    assembly.fasta

Prokka vs Bakta

Parsing Annotations with Python

Goal: Load Bakta/Prokka GFF3 output into a queryable database to extract CDS features and compute annotation quality metrics like coding density.

Approach: Create a gffutils in-memory database from the GFF3 file, iterate CDS features to extract locus tags and product names, and calculate coding density as total CDS bp divided by genome length.

import gffutils

def load_annotation(gff_file):
    '''Load GFF3 into a queryable database.'''
    db = gffutils.create_db(gff_file, ':memory:', merge_strategy='merge')
    return db

def extract_cds_features(db):
    '''Extract all CDS features with product annotations.'''
    features = []
    for cds in db.features_of_type('CDS'):
        features.append({
            'id': cds.id,
            'seqid': cds.seqid,
            'start': cds.start,
            'end': cds.end,
            'strand': cds.strand,
            'product': cds.attributes.get('product', ['unknown'])[0],
            'locus_tag': cds.attributes.get('locus_tag', [''])[0]
        })
    return features

def compute_coding_density(db, genome_length):
    '''Compute fraction of genome encoding proteins.

    Typical prokaryotic coding density: 85-95%.
    Values below 80% may indicate pseudogenes or annotation gaps.
    Values above 95% may indicate overlapping annotations.
    '''
    coding_bp = sum(cds.end - cds.start + 1 for cds in db.features_of_type('CDS'))
    return coding_bp / genome_length

db = load_annotation('bakta_out/my_genome.gff3')
cds_features = extract_cds_features(db)
print(f'Total CDSs: {len(cds_features)}')

Annotation QC

Expected Metrics by Genome Size

QC Checks

# Count annotated features
grep -c $'\tCDS\t' bakta_out/my_genome.gff3
grep -c $'\ttRNA\t' bakta_out/my_genome.gff3
grep -c $'\trRNA\t' bakta_out/my_genome.gff3

# Check for hypothetical proteins (ideally <40% of total CDSs)
grep -c 'hypothetical protein' bakta_out/my_genome.tsv

BUSCO on Predicted Proteins

busco -i bakta_out/my_genome.faa -m proteins -l bacteria_odb10 -o busco_proteins

Troubleshooting

Low Gene Count

• Check assembly completeness with BUSCO (genome mode)
• Verify correct translation table (--translation-table 4 for Mycoplasma)
• Inspect minimum contig length filter

Many Hypothetical Proteins

• Normal for novel organisms (30-50% is common)
• Try running InterProScan on the .faa file for additional annotations
• Consider eggNOG-mapper for orthology-based functional assignment

NCBI Submission

• Use --compliant flag for NCBI-ready output
• Ensure locus tags follow NCBI format (3-12 uppercase alphanumeric)
• Review .tsv output for annotation warnings

Related Skills

• functional-annotation - Add GO/KEGG/Pfam to predicted proteins
• ncrna-annotation - Detailed ncRNA identification with Infernal
• genome-assembly/assembly-qc - Assess assembly quality before annotation
• genome-intervals/gtf-gff-handling - Parse and manipulate GFF3 output