Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
ANNOVAR (ANNOtate VARiation) is an established command-line bioinformatics software package developed by Kai Wang and colleagues to perform functional annotation of genetic variants identified from high-throughput sequencing data. Operating directly on variant call files from whole-genome sequencing (WGS), whole-exome sequencing (WES), and targeted gene panels, ANNOVAR categorizes single nucleotide variants (SNVs), insertions and deletions (indels), and copy number variations across gene-based, region-based, and filter-based annotation protocols.
What it is used for in research
In modern genomics research, high-throughput sequencing instruments identify hundreds of thousands to millions of raw variant calls across sequenced samples. Transforming raw genomic coordinates into biologically meaningful insights requires mapping variants against gene models, conservation metrics, disease associations, and population allele frequencies. Researchers use ANNOVAR to systematically prioritize candidate disease mutations, characterize somatic cancer mutations, and streamline variant filtering pipelines.
ANNOVAR structures variant analysis across three core operational models:
- Gene-based annotation: Maps variants directly to gene models to evaluate their coding impact. Supported gene sets include NCBI RefSeq (
refGene), UCSC Genes (knownGene), Ensembl (ensGene), and GENCODE. The software categorizes variants into precise functional consequence classes, including synonymous SNVs, nonsynonymous SNVs, frameshift insertions or deletions, nonframeshift indels, stopgain mutations, stoploss mutations, splice-site disruptions, untranslated region (5' UTR and 3' UTR) alterations, intronic variants, and intergenic variants. - Region-based annotation: Identifies whether variants fall within defined functional, regulatory, or structural genomic segments. Typical targets include cytogenetic bands (
cytoBand), segmental duplications (genomicSuperDups), phylogenetically conserved elements across vertebrate species (such asphastConsElements), transcription factor binding sites, and ENCODE candidate cis-regulatory elements (cCREs). - Filter-based annotation: Evaluates variants against public population frequency repositories and curated disease databases. It cross-references allele frequencies from the Genome Aggregation Database (gnomAD exomes and genomes), the 1000 Genomes Project, and ExAC, alongside clinical annotations from ClinVar, dbSNP, and COSMIC. Furthermore, ANNOVAR compiles in silico pathogenicity prediction scores from the dbNSFP repository, incorporating algorithms such as SIFT, PolyPhen-2, CADD, REVEL, and MutationTaster.
Downstream research pipelines regularly feed ANNOVAR-annotated output into downstream statistical genetics workflows, such as utilizing maftools for downstream mutation analysis in cancer cohorts, benchmarking somatic calls against ACMG/AMP variant classification guidelines, or structuring computational analyses according to FAIR data principles.
How to install or access it
ANNOVAR is distributed exclusively through the official web portal maintained by the Kai Wang laboratory at https://annovar.openbioinformatics.org/. The software package is implemented in Perl and does not require complex binary compilation, making it portable across Unix, Linux, macOS, and Windows (via Windows Subsystem for Linux or Cygwin) environments.
Licensing and Registration Requirements
ANNOVAR is freely available for non-commercial academic research and non-profit educational purposes. Commercial organizations or users intending to integrate ANNOVAR into proprietary commercial services must obtain an explicit commercial license from the Center for Applied Genomics at the Children's Hospital of Philadelphia (CHOP) and the University of Southern California (USC). Because access is gated by license agreement acceptance, the software package cannot be retrieved via unauthenticated automated curl or wget commands; researchers must complete the registration form on the official download page to receive a personalized download link via email.
Prerequisites and Unpacking
To run ANNOVAR locally, ensure the computational environment satisfies the following system prerequisites:
- Perl Interpreter: Perl version 5.8 or higher (Perl 5.16+ recommended). Core modules such as
Pod::UsageandGetopt::Longmust be installed. - Disk Storage: The software script archive requires under 5 MB of disk space. However, precompiled reference databases require dedicated local disk storage: from 15 to 30 GB for standard exome filtering suites up to 200 GB or more for comprehensive whole-genome resources (such as whole-genome gnomAD and full dbNSFP tables).
Once downloaded, extract the compressed tarball in the chosen working directory:
# Extract the official archive
tar -zxvf annovar.latest.tar.gz
# Navigate into the unpacked directory
cd annovar/
ls -F
# annotate_variation.pl* convert2annovar.pl* table_annovar.pl* example/ humandb/
For researchers seeking an interactive graphical interface or small-scale variant triage without command-line execution, the official wANNOVAR web server (http://wannovar.wglab.org/) offers an alternative hosted interface incorporating automated phenotype-driven variant prioritization.
Step-by-step instructions
The standard, authoritative workflow for annotating human genetic variants involves four primary steps: downloading reference database files, preparing input variant coordinates, running the consolidated table_annovar.pl pipeline, and applying clinical frequency filters.
Step 1: Download Required Reference Databases
ANNOVAR uses the annotate_variation.pl utility with the -downdb and -webfrom annovar parameters to fetch precompiled, indexed databases directly into a local storage folder (conventionally named humandb/). The example commands below fetch resources for the human reference genome build hg38 (or replace with hg19 for legacy builds):
# Create directory for local reference databases
mkdir -p humandb/
# 1. Download gene definitions (NCBI RefSeq)
perl annotate_variation.pl -buildver hg38 -downdb -webfrom annovar refGene humandb/
# 2. Download regional annotations (cytogenetic bands and segmental duplications)
perl annotate_variation.pl -buildver hg38 -downdb -webfrom annovar cytoBand humandb/
perl annotate_variation.pl -buildver hg38 -downdb -webfrom annovar genomicSuperDups humandb/
# 3. Download population allele frequency databases (gnomAD exomes and dbSNP)
perl annotate_variation.pl -buildver hg38 -downdb -webfrom annovar gnomad41_exome humandb/
perl annotate_variation.pl -buildver hg38 -downdb -webfrom annovar avsnp151 humandb/
# 4. Download clinical and functional prediction databases (ClinVar and dbNSFP)
perl annotate_variation.pl -buildver hg38 -downdb -webfrom annovar clinvar_20240917 humandb/
perl annotate_variation.pl -buildver hg38 -downdb -webfrom annovar dbnsfp47a humandb/
Step 2: Prepare Input Variant Data
ANNOVAR supports two primary input formats: native VCF files and the native five-column .avinput text format. When starting from standard multi-sample or single-sample VCF files, best practice in computational genomics requires normalizing and splitting multi-allelic records first using standard tools before passing to ANNOVAR. For detailed guidance on VCF field syntax, consult the guide on reading and interpreting VCF files, and see the breakdown of gVCF and standard VCF differences.
Alternatively, external formats can be converted to the native five-column format (Chromosome, Start, End, Reference Allele, Alternative Allele) using convert2annovar.pl:
# Convert standard VCF format into ANNOVAR input format
perl convert2annovar.pl -format vcf4 input_variants.vcf > input_variants.avinput
Step 3: Execute table_annovar.pl Consolidated Annotation
Modern pipelines should use table_annovar.pl, which executes all requested gene-, region-, and filter-based protocols in a single pass and synthesizes findings into an aligned tab-delimited text summary and an annotated VCF:
perl table_annovar.pl input_variants.vcf humandb/
-buildver hg38
-out sample_cohort
-remove
-protocol refGene,cytoBand,genomicSuperDups,gnomad41_exome,clinvar_20240917,avsnp151,dbnsfp47a
-operation g,r,r,f,f,f,f
-nastring .
-vcfinput
-polish
-otherinfo
Parameter definitions for this authoritative command:
input_variants.vcf: The path to the query variant file.humandb/: Directory path housing the downloaded reference databases.-buildver hg38: Designates reference assembly version (must match database builds).-out sample_cohort: Base prefix for generated output files.-remove: Deletes intermediate working files upon successful completion to save disk storage.-protocol: Comma-delimited list of database names to evaluate in order.-operation: Comma-delimited list specifying the operational mode for each respective protocol in-protocol(g= gene-based,r= region-based,f= filter-based).-nastring .: Sets the character used when a variant has no record in a database (period is standard in VCF format).-vcfinput: Instructs ANNOVAR that the input file is in VCF format and to produce an annotated VCF (*.multianno.vcf).-polish: Refines transcript coordinate representation and exon numbering for splicing and indels.-otherinfo: Retains original VCF INFO, FORMAT, and sample genotype fields in the tab-delimited output.
Step 4: Prioritize Candidate Variants
Once annotation finishes, researchers filter the resulting tabular output to isolate pathogenic candidates according to established filtering cascades:
- Functional consequence filter: Retain variants where
Func.refGeneisexonicorsplicing, discarding synonymous coding changes unless splice-disruption predictions are severe. - Population frequency filter: Exclude common benign polymorphisms by filtering on
gnomad41_exome_AF < 0.01(or< 0.001for dominant mendelian phenotypes). - Pathogenicity filter: Query
CLNSIGfor ClinVar pathogenic designations and filter on in silico prediction thresholds (e.g., CADD Phred score ≥ 20, indicating top 1% most deleterious variants across the genome).
Output and how to read it
The primary output generated by table_annovar.pl is the tab-delimited text file sample_cohort.hg38_multianno.txt (alongside sample_cohort.hg38_multianno.vcf when -vcfinput is enabled). Each line in the tabular output corresponds to a specific genomic variant, with columns representing the requested annotation protocols.
Key columns and how to interpret their contents:
Chr,Start,End,Ref,Alt: Core variant coordinates and alleles. ANNOVAR uses 1-based inclusive genomic coordinates. For single nucleotide substitutions,StartequalsEnd. For deletions,StartandEnddelineate deleted bases, withAltreported as-. For insertions,StartandEnddelineate flanking bases, withRefreported as-.Func.refGene: The genomic locus category relative to the nearest transcript, such asexonic,splicing,intronic,UTR5,UTR3,upstream,downstream, orintergenic.Gene.refGene: Official HGNC gene symbol(s) associated with the transcript. When overlapping multiple genes, symbols are delimited by commas.GeneDetail.refGene: Supplementary transcript details, including distances to transcription start or stop sites for non-coding variants, or transcript IDs for splice-junction annotations.ExonicFunc.refGene: Detailed functional consequence for exonic variants. Values includenonsynonymous SNV,synonymous SNV,frameshift insertion,frameshift deletion,nonframeshift insertion,nonframeshift deletion,stopgain(premature termination codon), andstoploss.AAChange.refGene: Formatted transcript mutation descriptor identifying the transcript identifier, exon number, cDNA alteration, and resulting protein alteration (for example,NM_000546:exon4:c.G215C:p.R72P).cytoBand: Regional chromosome band mapping (e.g.,17p13.1).genomicSuperDups: Segmental duplication score indicating sequence identity fraction. High values (≥0.95) alert researchers that the variant falls in a low-complexity or duplicated region prone to alignment artifacts.gnomad41_exome_AF: Global allele frequency in the gnomAD exome cohort. A period (.) denotes that the variant was absent from the reference cohort, indicating potential novelty.CLNSIG: Curated clinical significance from ClinVar (e.g.,Pathogenic,Likely_pathogenic,Benign, orUncertain_significance).avsnp151: Reference SNP cluster identifier (rsID) mapped from dbSNP.SIFT_score,Polyphen2_HDIV_score,CADD_phred: Quantitative in silico deleteriousness metrics from dbNSFP.
Troubleshooting
Bioinformatics practitioners encounter several recurrent operational challenges when integrating ANNOVAR into computational pipelines. The matrix below details causes and concrete technical remedies:
| Issue and Symptom | Root Cause | Resolution and Recommended Fix |
|---|---|---|
| Chromosome notation mismatch (Annotations appear completely empty or missing across all rows) | Discrepancy between chromosome naming conventions in the input VCF (e.g., UCSC style chr1, chr2) and the database index files (e.g., Ensembl/NCBI style 1, 2), or vice versa. |
Inspect the first column of database files in humandb/. Standardize the input VCF before running ANNOVAR using bcftools annotate --rename-chrs chrom_map.txt or use sed to strip or append the chr prefix so coordinates match database indexes exactly. |
| Multi-allelic VCF site corruption (Second alternate allele skipped or misaligned columns) | VCF files containing multi-allelic records (e.g., ALT=A,C) on a single row are not decomposed prior to annotation. |
Decompose and left-align variants before running ANNOVAR using bcftools norm -m -any -f ref.fasta input.vcf -o normalized.vcf. This ensures each alternate allele is evaluated on its own discrete record. |
Database download failure (annotate_variation.pl -downdb fails with HTTP 404 or connection timeout) |
Upstream repository version deprecation or mirror server throttling on the ANNOVAR hosting infrastructure. | Confirm available database identifiers on the official ANNOVAR website. Alternatively, download current raw files directly from original sources (such as NCBI ClinVar FTP or gnomAD public Google Cloud buckets) and reformat using ANNOVAR auxiliary scripts. |
Missing Perl module errors (Execution halts with Can't locate Pod/Usage.pm in @INC) |
Minimal Linux distributions or stripped container images (e.g., minimal Ubuntu, Debian slim, Alpine) package Perl without standard core documentation utilities. | Install missing core packages via the system package manager: run apt-get install -y perl-doc on Debian/Ubuntu or cpanm Pod::Usage via CPAN. |
| Excessive disk consumption (Scratch storage fills rapidly during large-cohort annotation) | Omitting intermediate file cleanup flags causes ANNOVAR to preserve multiple multi-gigabyte temporary step files for each protocol. | Always supply the -remove parameter in table_annovar.pl to delete temporary protocol step files immediately upon integration into the master table. |
Free alternatives
Multiple open-source and freely accessible variant annotation engines exist across the bioinformatics ecosystem. The comparative table below outlines functional characteristics across leading software tools:
| Software Tool | Developer / Institution | Runtime Engine | License & Commercial Terms | Primary Input / Output Formats | Key Strengths |
|---|---|---|---|---|---|
| ANNOVAR | Kai Wang Lab (USC / CHOP) | Perl (CLI) | Free for non-commercial academic research; commercial license required for industrial use | VCF, .avinput / Tabular text (.multianno.txt), annotated VCF | Lightweight footprint, standalone execution without server daemon, intuitive tab-delimited output |
| Ensembl VEP (Variant Effect Predictor) | EMBL-EBI | Perl, C, REST API, Docker | Open Source (Apache 2.0; free for all academic and commercial use) | VCF, HGVS identifiers / VCF, JSON, Tabular | Strict Sequence Ontology (SO) compliance, comprehensive plugin ecosystem, native transcript cache indexing |
| SnpEff & SnpSift | Pablo Cingolani | Java (JVM) | Open Source (GNU LGPLv3; free for all users) | VCF / Annotated VCF, HTML summary report | Extremely fast high-throughput execution speed, powerful integrated variant filtering syntax via SnpSift |
| OpenCRAVAT | Karchin Lab (Johns Hopkins University) | Python (CLI & Web GUI) | Open Source (Apache 2.0; free for all users) | VCF / Interactive Web GUI, SQLite, TSV, VCF | Modular app store architecture with over 100 annotation modules, dynamic visual filtering interface |
| Nirvana | Illumina Inc. | C# (.NET Core) | Open Source (Apache 2.0 / MIT components) | VCF / Structured JSON | Engineered specifically for clinical high-throughput WGS production pipelines, sub-second execution on large cohorts |
FAQ
What is the difference between table_annovar.pl and annotate_variation.pl?
annotate_variation.pl is ANNOVAR's underlying core execution engine. It was originally designed to run one annotation protocol at a time and generate separate intermediate output files. table_annovar.pl is the modern wrapper script that coordinates multiple gene-, region-, and filter-based annotations in a single command, automatically managing input conversions and compiling all results into a single unified tab-delimited table and annotated VCF file.
Can ANNOVAR annotate both hg19 and hg38 human genome builds?
Yes. ANNOVAR fully supports both hg19 (GRCh37) and hg38 (GRCh38), as well as earlier human builds (hg18) and non-human model organisms (mouse, fly, worm, zebrafish). When running commands, the -buildver argument must explicitly specify the genome version (for example, -buildver hg38), and all queried databases in humandb/ must have been fetched with the identical build flag.
Is ANNOVAR free for commercial or industrial bioinformatics pipelines?
No. While ANNOVAR is free for non-profit academic research, commercial entities and developers incorporating ANNOVAR into revenue-generating diagnostics, software, or services are legally required to obtain a commercial license through the University of Southern California and Children's Hospital of Philadelphia.
How does ANNOVAR define exonic variant consequences compared to Sequence Ontology terms?
ANNOVAR uses descriptive clinical nomenclature (such as nonsynonymous SNV, synonymous SNV, stopgain, and frameshift insertion) rather than strict Sequence Ontology (SO) terms (such as missense_variant or stop_gained). In contrast, tools like Ensembl VEP adhere strictly to standardized Sequence Ontology terms.
Can ANNOVAR process copy number variants (CNVs) and structural variants?
Yes. ANNOVAR can evaluate copy number variants and structural variations when supplied in the five-column .avinput format. For structural deletions or duplications, users specify the chromosome, start boundary, end boundary, 0 as the reference allele, and the copy number call or event descriptor (such as <CNV>) in the alternative column.
What does a dot (.) mean in the output columns of multianno.txt?
When the -nastring . parameter is specified during execution, a dot denotes a missing value, indicating that the variant coordinate was not present in that particular reference database. For population frequency databases like gnomAD, a dot signifies that the variant was unobserved in the reference cohort.
Sources
- ANNOVAR Documentation Portal — Official software manual, documentation, and user guides (Wang Lab, read 2026-11-10).
- ANNOVAR Download & Installation Guide — Software licensing terms, registration portal, and installation prerequisites (Wang Lab, read 2026-11-10).
- Wang, K., Li, M., & Hakonarson, H. (2010). ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Research, 38(16), e164. doi:10.1093/nar/gkq603 (Read 2026-11-10).
- Yang, H., & Wang, K. (2015). Genomic variant annotation and prioritization with ANNOVAR and wANNOVAR. Nature Protocols, 10(10), 1556–1566. doi:10.1038/nprot.2015.105 (Read 2026-11-10).
- McLaren, W., et al. (2016). The Ensembl Variant Effect Predictor. Genome Biology, 17(1), 122. doi:10.1186/s13059-016-0974-4 (Read 2026-11-10).
- Cingolani, P., et al. (2012). A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff. Fly, 6(2), 80–92. doi:10.4161/fly.19695 (Read 2026-11-10).







