Ncbi Genome Annotation
NCBI genome annotation is the computational process of identifying gene locations, coding regions, regulatory elements, and other functional features on a submitted genome assembly. This guide is written for researchers, bioinformaticians, and genome submitters who need a practical, source bounded understanding of how NCBI handles genome annotation from raw assembly to public RefSeq record.
If you are planning to submit a genome to GenBank or want to understand what annotation means on an NCBI record, start here. The annotation pipeline combines ab initio gene prediction with evidence from transcript sequences and protein homology. The output is a standardized annotation that appears in RefSeq and GenBank databases.
NCBI Bookshelf provides the authoritative technical reference for the annotation pipeline, while the Galaxy Training Network offers hands on workflows for users who want to run their own annotation before submission.
At a Glance
| Aspect | Details |
|---|---|
| What it does | Identifies genes, transcripts, proteins, and non coding features on a genome assembly |
| Input | A complete or draft genome assembly in FASTA format |
| Output | A GenBank or RefSeq annotation file with coordinates and functional names |
| Pipeline type | Eukaryotic (Eukaryotic Genome Annotation Pipeline) or Prokaryotic (Prokaryotic Genome Annotation Pipeline) |
| Evidence used | Transcript alignments (RNA seq, EST), protein homology, ab initio models |
| Decision points | Organism type, assembly quality, submission pathway (GenBank vs RefSeq) |
| Key limitation | Annotation is a prediction, experimental validation is required for high confidence |
| Typical timeline | 2 6 weeks for a standard genome after submission |
Core Concepts and Decision Points
Annotation vs. Assembly
An assembly is a reconstructed string of nucleotides. Annotation makes sense of that string by marking where genes start and stop, where introns are spliced, and which RNA or protein products are produced. Without annotation, a genome is just a long sequence.
Prokaryotic vs. Eukaryotic Pipelines
NCBI runs two separate annotation pipelines. The Prokaryotic Genome Annotation Pipeline (PGAP) handles bacterial and archaeal genomes. It uses a set of HMM models and curated reference protein databases. The Eukaryotic Genome Annotation Pipeline (EGAP) handles everything else from fungi to plants to mammals. It relies heavily on transcript evidence and comparative protein data.
EMBL EBI Training offers parallel resources for understanding what these pipelines do at the European level, but NCBI maintains its own distinct workflows.
GenBank vs. RefSeq Annotation
When you submit a genome through GenBank, the annotation is often provided by the submitter or computed by NCBI as a service. RefSeq annotation is a curated, standardized version that NCBI produces for reference genomes. Not every assembly gets a full RefSeq annotation, only those selected as reference genomes or submitted through the RefSeq initiative.
The key decision is whether you want a quick GenBank annotation for a new isolate or a full RefSeq annotation for a model organism. Bioconductor provides packages to compare and contrast these annotation sets programmatically.
Evidence Types and Their Role
The pipeline uses three main evidence streams:
- Transcript alignments RNA seq reads, ESTs, or full length cDNAs aligned to the genome.
- Protein homology known proteins from related species aligned to the genome.
- Ab initio prediction computational gene finders trained on the evidence.
Better transcript evidence produces more accurate exon intron boundaries. For non model organisms, homology based annotation is often the fallback.
Practical Workflow
Step 1: Prepare Your Assembly
Your assembly must meet NCBI quality standards. The assembly should be in a single FASTA file, with no ambiguous bases beyond the expected levels. For eukaryotes, chromosome level or scaffold level assemblies are preferred. For prokaryotes, a complete circular chromosome is ideal but not required.
Use the NCBI Sequence Read Archive to deposit raw sequencing reads that support your assembly. These reads can later be used as transcript evidence if you submit RNA seq data.
Step 2: Submit to GenBank or RefSeq
Go to the NCBI Submission Portal. You can submit as a simple genome (GenBank) or request RefSeq annotation. The submission form asks for metadata: organism name, sequencing technology, assembly method.
After submission, NCBI runs the appropriate pipeline. You will receive a draft annotation that you can review before release.
Step 3: Review Annotation Quality
NCBI provides an annotation report with statistics:
- Number of genes
- Number of coding sequences (CDS)
- Number of non coding RNAs
- Coverage of known proteins
- Percentage of genome annotated
Check these numbers against expectations for your organism. For a typical bacterium, expect 4,000 to 6,000 protein coding genes. For a mammal, expect 20,000 to 25,000.
Step 4: Download and Use the Annotation
Once released, you can download annotation files in GFF3, GenBank flat file, or FASTA format for proteins and transcripts. These are available through the NCBI FTP site and the Assembly database.
Global Landscape of Publicly Available Human Oral Microbiome Data shows a real world example where NCBI annotation was used to characterize microbial genomes from oral samples. The study relied on the standardized annotation to compare gene content across isolates.
Quality Checks
What to Look For
- Contamination: Ensure no vectors, adapters, or foreign DNA. NCBI screens for contamination, but you should also verify. Use tools like CheckM or BUSCO on your assembly before submission.
- Frameshifts: Disrupted open reading frames can indicate sequencing errors or pseudogenes. The annotation pipeline flags these.
- Missing functional domains: Compare your annotation to closely related species. If key metabolic genes are absent, the annotation may be incomplete.
- Gene density: Unusually high or low numbers of genes per megabase can signal annotation errors or assembly problems.
The first complete mitochondrial genome sequences of Neopsylla specialis dechingensis demonstrates how careful quality checks on mitochondrial annotation can resolve phylogenetic placement. The authors performed manual curation after the NCBI pipeline to verify gene boundaries.
Cross Validation
If you have independent transcript data (RNA seq), align it to your annotated genome. Check whether annotated exons align to actual reads. The Enhanced RNA Seq Expression Profiling in Non model Organisms Using Custom Annotations protocol describes how to validate and improve NCBI annotation using your own RNA evidence.
Common Mistakes
Mistake 1: Submitting a Poor Quality Assembly
If your assembly has high N50 but many misassemblies, the annotation will place genes in wrong locations or miss them entirely. Always run assembly quality metrics before submission.
Mistake 2: Ignoring Taxonomy
NCBI uses taxonomy to select reference proteins. If you identify your organism incorrectly or use a genus level name when a species level name is warranted, the pipeline may use distant homologs and produce poor annotation.
Mistake 3: Not Providing Evidence When Possible
Even if the pipeline can run without transcript evidence, providing RNA seq data dramatically improves accuracy. Do not skip this step if you have the data.
Mistake 4: Assuming Annotation Equals Truth
NCBI annotation is computational. It will miss alternatively spliced isoforms, non coding RNAs that are not well conserved, and genes with unusual structure. Always treat genome annotation as a hypothesis.
Whole genome sequence of a superbug Escherichia coli strain KAB AI 497 illustrates how resistance gene annotation from NCBI can guide further lab validation. The authors confirmed the annotation by PCR and phenotypic testing because automated annotation can miss novel variants.
Limits of Interpretation
Annotation is a probabilistic prediction. The boundaries between true genes and pseudogenes, between real transcripts and artifacts, are not sharp. Here are specific limits:
- Pseudogenes: The pipeline may annotate a pseudogene as a coding gene if it still has some sequence similarity. Manual curation is needed.
- Non coding RNA: Small nucleolar RNAs, microRNAs, and long non coding RNAs are poorly predicted without direct evidence.
- Alternative splicing: For eukaryotes, the pipeline captures major isoforms but rarely all isoforms.
- Novel genes: Truly novel genes with no homology to any known sequence are almost always missed.
- Frameshifts and sequencing errors: These can produce false gene fusions or truncated proteins.
Genomic characterisation of a high level azithromycin resistant Neisseria gonorrhoeae and Whole genome sequence data of a multidrug resistant Klebsiella pneumoniae both used NCBI annotation as a starting point but supplemented with manual inspection of resistance genes. That combination of automated annotation and expert review is the gold standard.
Frequently Asked Questions
Q1: Can I annotate a genome myself before submitting to NCBI?
Yes. You can run PGAP locally or use Galaxy to run annotation tools. NCBI accepts annotations provided by submitters. However, the NCBI pipeline is often preferred for standardization.
Q2: How long does NCBI take to annotate a genome?
Typical turnaround is 2 to 6 weeks for a bacterial genome. Eukaryotic genomes with transcript evidence can take 4 to 8 weeks. The pipeline runs in batch and priority is given to RefSeq submissions.
Q3: What if my organism has no close relative in the database?
The pipeline uses protein families (Pfam, TIGRFAM) and can still annotate many genes. However, you will get fewer named functional annotations. You may need to provide RNA seq data to improve gene model accuracy.
Q4: Can I update an annotation after it is released?
Yes. You can submit a revised assembly or new evidence that triggers a re annotation. RefSeq records are periodically updated, but GenBank submissions can be updated by the original submitter.
References and Further Reading
- NCBI Bookshelf (authoritative reference on annotation pipelines)
- EMBL EBI Training (complementary training on annotation and bioinformatics)
- Galaxy Training Network (workflows for genome annotation)
- Bioconductor (R packages for manipulating and comparing annotations)
- NCBI Sequence Read Archive (deposit raw sequencing data to support annotation)
- Global Landscape of Publicly Available Human Oral Microbiome Data (case study using NCBI annotation)
- The first complete mitochondrial genome sequences of Neopsylla specialis dechingensis (mitochondrial annotation and curation)
- Enhanced RNA Seq Expression Profiling in Non model Organisms Using Custom Annotations (protocol for validating NCBI annotation)
- Whole genome sequence of a superbug Escherichia coli strain KAB AI 497 (genome annotation with resistance gene validation)
- Genomic characterisation of a high level azithromycin resistant Neisseria gonorrhoeae (annotation and phylogenomics)