Due to staff annual leave, an increase in the response time for helpdesk tickets is expected between the 24th December and 2nd January. Please refrain from creating duplicate tickets during this time; we will action them as soon as possible.
We apologise for any inconvenience caused during this time, and thank you for your patience and understanding.
The European Variation Archive is an open-access database of all types of genetic variation data from all species.
All users can download data from any study, or submit their own data to the archive. You can also query all variants in the EVA by study, gene, chromosomal location or dbSNP identifier using our Variant Browser.
We will be adding new features to the EVA on a regular basis, and welcome your comments and feedback.
Search for SNPs
Short genetic variants studies (<50bp)
Structural variants studies (>50bp)
Alternative resources for data not accepted by EVA
- Submit structural variations that cannot be expressed in VCF(s) to DGVa.
- Submit variations with sensitive clinical data to EGA.
- Submit variations with clinically relevant genetic variant data, i.e. data that relates genetic variation(s) with clinical significance values (e.g. pathogenic, benign, etc.), to the ClinVar archive at NCBI.
Key stages of EVA submissions
- Prepare valid VCF file(s), which can be validated prior to submission using the EVA VCF validation suite.
- Complete a metadata template describing the samples and analyses in your study. Please provide as much metadata as possible since this information is extremely useful for downstream analysis and is directly related to the frequency at which datasets archived at EVA are reused. For reference, here is an example of a completed metadata template.
Please also note that the template requires the submitter to fill in some personal data, which will be used as described in our privacy notice.
Contact firstname.lastname@example.org to request a submission. You will receive a custom private FTP account to deposit your data.
Upload your VCF file(s), metadata template and any associated data file(s) to your private FTP location.
If you have any questions related to the European Variation Archive resource, please contact us.
Follow us on Twitter using @evarchive
The general structure of a EVA REST web service URL is:
* version: indicates the version of the API, this defines the available filters and JSON schema to be returned. Currently there is only version 'v1'.
* category: this defines what objects we want to query. Currently there are five different categories: variants, segments, genes, files and studies.
* resource: specifies the resource to be returned, therefore the JSON data model.
* filters: each specific endpoint allows different filters.
REST web services have been implemented using GET protocol since only queries are allowed so far. Several IDs can be concatenated using comma as separator.
For more detailed information about the API and filters you can visit the project wiki and Swagger documentation.
Some example of queries include:
* To fetch all the variants in a segment region:
* To fetch all the info of a variant:
* To fetch all the Short Genetics Variations studies:
* To fetch all the Structural Variations studies:
* To fetch all info of a study:
* To fetch all file information of a study:
What is the European Variation Archive (EVA)?
The European Variation Archive (EVA) is EMBL-EBI's open-access genetic variation archive. The EVA accepts submission of all types of precise genetic variants, ranging from single nucleotide polymorphisms to large structural variants, observed in germline or somatic sources, from any organism. The EVA permits access to these data at two distinct levels:
i) The raw variant data as was submitted to the EVA, via the EVA Study Browser
What are the EVA normalisation and variant processing steps?
Variants submitted to the EVA have been determined by a number of different algorithms and software packages. As a result, the VCF files generated by these differing methodologies describe variants in a number of different ways. The primary processing step of the EVA is to normalise variant representation following two basic rules:
- Each variant is shifted to be left-aligned
- The Start and End positions represent exactly the range where the variation occurs (which could, in the case of insertions, result in the reference allele being recorded as 'empty')
Examples of our variant normalisation process can be seen here
Once variants have been normalised, the EVA uses the Variant Effect Predictor (VEP) of Ensembl to annotate variant consequences. The variant consequences are described using Sequence Ontology terms and both the VEP version and Ensembl gene build used are described via the "i" help bubbles on the EVA Variant Browser.
N.B. Variants that have been mapped to a reference genome sequence that is not supported by Ensembl are not annotated.
The EVA adopts the classical definition of allele frequency (AF): 'a measure of the relative frequency of an allele at a genetic locus in a given population'. The AF value(s) stored by the EVA for each variant is (are) study specific - i.e. the same variant reported in two distinct studies shall be given two allele frequencies, one for each study. There are two methodologies by which the EVA is able to determine allele frequency values, dependent on the datatype of the study in question:
Variants associated with genotypes:
For variants associated with genotypes, the EVA determines the AF values via the calculation:
AF = (number of alternate allele observations (AC)) / (number of observations (AN))
The result of this calculation allows the EVA to also store the minor allele frequency (MAF) for each variant (defined as the minimum of the reference or alternative allele frequency) and the MAF allele (the allele associated with the MAF).
Variants not associated with genotypes:
For variants that are not associated with genotypes, the EVA is dependent on the AF value(s) estimated from the primary data and provided in the submitted VCF files(s). AF values that are specifically provided in the submitted aggregated VCF file(s) are directly stored. In cases where no AF is provided then the EVA uses the AC and AN values in the submitted aggregated VCF file(s) to calculate AF value(s) via the calculation:
AF = AC / AN
Population / sample cohort allele frequency values:
The EVA accepts submission of pedigree files, or structured samples (using "derived_from" and/or "subject" layers), to define populations and cohorts within studies. In cases where such information is associated with variants that have genotypes then the EVA calculates intra-study population/cohort specific AF values via the method described above, with the caveat that the (total number of populations/cohorts):(total number of samples) ratio must be less than 1:10. For studies that do not contain genotypes but instead provide intra-study population/cohort AF values in the submitted aggregated VCF file(s), or AC and AN values, then these are directly stored, or calculated by the EVA using the method described above, again with the caveat that a ratio of 1:10 (total number of populations/cohorts):(total number of samples) must not be exceeded.
*NB: there are a low number of variants for which the EVA is unable to determine any allele frequency value(s) as the submitted VCF file(s) contain neither genotypes nor AF or AC and AN values. The EVA discourages submission of variants that cannot be associated with an AF.
With whom does the EVA collaborate?
The EVA & GEUVADIS European Exome Variant Server
The EVA & GEUVADIS European Exome Variant Server (GEEVS;) work in collaboration to coordinate common data formats for data exchange. As part of this collaboration, we fully endorse the variant calling protocol detailed on the GEEVS website as adherence to this protocol for variant calling permits direct comparison and/or aggregation of results from different datasets.
The EVA & Príncipe Felipe Research Center
Some of the technical and analytical features of the EVA were developed in collaboration with the department of Computational Genomics led by Joaquin Dopazo at the Principe Felipe Research Centre Computational Genomics Department (CIPF).
EVA & The Genomics and Bioinformatics Platform of Andalusia
Early development of the EVA was carried out in collaboration with the Bioinformatics Department at the Genomics and Bioinformatics Platform of Andalusia.
How can I follow the development of the EVA?
The following are our GitHub repositories:
The EVA VCF validator checks that a file is compliant with the VCF specification. It includes and expands the validations supported by the vcftools suite. It supports versions 4.1, 4.2 and 4.3 of the specification.
The EVA pipeline processes VCF files, stores the variation data in a database and post-processes it, in a way that can be later consumed via web services.
The EVA web services serve the data generated and stored by the EVA pipeline. They follow the REST paradigm and can be consumed by any external application.
The EVA website displays the data served by the EVA REST web services API in a user-friendly way.
We would like to acknowledge the following software support.
How can I consume variant data from the EVA?
Links are shown on the EVA Study Browser to both the raw submitted files and the 'EVA browsable files'. Not all submitted VCF files for a study are browsable due to overlapping information. For example, we have studies where the submitted VCF files contain the sample and variant data split by population, but also merged together. It would be redundant for the EVA to load all of these data to our variant warehouse. Please note that due to processing time required not all studies have 'EVA browsable files' just yet.
Which browsers does the EVA website support?
The EVA website employs HTML5 technologies and standards. Chrome (version 18 or higher), Firefox (version 12 or higher), Safari (version 6 or higher) , Opera (version 12 or higher) and Internet Explorer (version 10 or higher) are fully supported, however older versions of these browsers may give rise to errors.
The EVA website also supports mobile versions of web browsers, with limited functionality. Please report all errors to the EVA helpdesk.
How can I search EVA with existing dbSNP identifiers?
Can I download accessioned variants from the EVA?
Yes. After an rs build the EVA will provide a VCF dump for each species/assembly combination. These files shall be in the same style as the current VCFs that dbSNP generates (i.e. contain the rs ID positional information, but not the genotypic nor annotation data). Furthermore we are developing a 'VCF Dumper' tool which shall allow users to generate a custom VCF file based on filtering options available at the EVA Variant Browser.
How can I track variants that have been generated, merged and/or deprecated over time?
In addition to outputting a VCF file for each rs build the EVA shall also generate a file that will allow users to track changes to variant numbers over time. These files shall include details of all newly generated identifiers for each rs build, a list of those variants that have been deprecated as well as information on any variants that have been merged together. The EVA shall follow the same rules as dbSNP for generating, deprecating and merging variant identifiers.
Why submit data to the EVA?
The EVA provides to the community a completely free, secure and permanent solution to data sharing. Each project, VCF file and sample that is submitted to the EVA is assigned a unique identifier that is accessible in perpetuity and is therefore able to be referenced in publication, for example. The EVA helpdesk provides support to submitters, and users, to ensure accurate represention and proper integration of the submitted data with other EMBL-EBI resources such as the EGA and Ensembl. A final advantage of submitting to the EVA is that variants are brokered to the National Centre for Biotechnology Information (NCBI) on the submitter's behalf, negating the requirement for independent submissions.
Is my data suitable for submission to the EVA?Is your data open access?
The most important consideration is that all data archived at the EVA is open access. As such, there are no restrictions as to who can access the data or how such data is reused. It is the submitter's responsibility to ensure that the data archived at the EVA complies to this open access policy.
If consent was gathered for any individual genotype data in the study then we will require a consent statement to be completed prior to submission.
Is your data represented in a valid VCF file?
Genetic variants submitted to the EVA must be described in the Variant Call Format (VCF). The EVA only accepts VCF files that adhere to the specification and match a reference sequence. Submitters can make use of the EVA VCF validation suite to ensure they fulfill these requirements. Furthermore, VCF files submitted to the EVA should provide either sample genotypes and/or aggregated sample summary-level allele frequencies.
- Population allele frequencies: Alleles typically exist at different frequencies in different populations and this information provides useful supporting evidence for all submitted variants. Population-level allele frequencies can be provided in the INFO field of the VCF (see section 1.6.1 of the VCF specification).
- Individual genotypes: EVA accepts individual genotypes from samples provided by donors that have consented to having their DNA sequence housed in a public database. Genotypes can be provided in the FORMAT field of the VCF (see section 1.6.2 of the VCF specification).
Is the reference sequence INSDC registered?
The EVA requires that all variants are submitted with an asserted position on an INSDC sequence. This means that the reference allele submitted for every variant must match the position in a sequence that has been accessioned in either GenBank or ENA database. An “INSDC sequence” can be any of, but not restricted to, the following assembled sequences:
- Genome assembly
- Gene sequence
Aligning all of our submitted data with INSDC sequences enables integration with other EMBL-EBI resources including Ensembl and is an important step in maintaining standardisation at the EVA. Importantly this requires all sequence identifiers in your VCF to match those in the reference FASTA file.
How do I generate a VCF file?
The basic template of a minimal VCF file looks like this (where no genotypes are given):
##fileformat=VCFv4.3 #CHROM POS ID REF ALT QUAL FILTER INFO
We advise submitters to consult the VCF specification (specifically sections 1.1 and 5) when converting data to VCF in order to ensure that files generated are valid, as this is a mandatory requirement.
Examples of valid VCF submissions
- PRJEB27443 - Simple INDEL submission with allele frequencies
- PRJEB27177 - Single sample VCF with genotype
- PRJEB27093 - Multiple sample VCF with genotypes
- PRJEB4163 - VCF with population allele frequencies and genotypes
- PRJEB6041 - VCF with allele counts (for calculating AF)
Further examples can be found on our FTP site.
Useful tools for handling VCF files
- VCF Validation: EVA VCF validation suite
- Sorting, merging and manipulation of VCF files: bcftools
- Conversion from genotyping arrays/PED files: PLINK1.9 (using --recode vcf and --reference-allele to select the reference alleles)
- Variant callers: GATK HaplotypeCaller, Platypus, VarScan, LoFreq, FreeBayes, SNVer, SAMtools and VarDict
What happens to my data once submitted?
Submission validation processes
The EVA VCF validation suite provides tools to check file validity previous to submission.
All VCF files submitted to EVA are validated for adherence to the format standard using the EVA VCF validation suite which includes all the checks from the vcftools suite, and some more that involve lexical, syntactic and semantic analysis of the VCF input. The EVA VCF validation suite also includes a debugging tool to automatically correct many of the common errors found in files. To improve processing time, submitters are encouraged to pre-validate VCF files prior to submission.
To improve interoperability of variant data submitted to EVA with other resources at EMBL-EBI, and the wider open-access community, all submitted VCF files are subject to validation against the INSDC accessioned reference sequence indicated by the submitter in their metadata. The EVA VCF validation suite provides an assembly checker tool to validate the reference alleles in the VCF files against the reference sequence. EVA is able to only accept files that match a known sequence at 100%.
All VCF files and novel samples that are submitted to EVA are permanently and securely archived at the European Nucleotide Archive and BioSamples, respectively. The EVA provides access to all submitted data via the Study Browser.
Variants within VCF files submitted to EVA are normalised, annotated and used for statistical calculations (via methodologies described below) and these EVA processed data are available via the Study Browser, Variant Browser and the REST API.
Do EVA and dbSNP accept submission of the same genetic variation file format?
Yes. EVA, like dbSNP, accepts submission of genetic variation data that is described in Variant Call Format (VCF) files. However, unlike dbSNP, EVA does not have a custom VCF specification. Instead EVA only accepts submission of VCF files that conform to the specification guidelines. Conformity to these guidelines is checked at time of submission, but can also be checked beforehand using the EVA VCF validation suite. Further guidelines for submission to EVA can be found here.
Do EVA and dbSNP collect metadata in the same way?
No. Although there are many similarities, the EVA and dbSNP metadata templates, which are submitted along with VCF files, are different. The EVA metadata template asks the submitter to provide information about the project and samples so that these can be archived simultaneously along with the VCF file(s), whereas this is a stepwise process during dbSNP submission. The EVA metadata template can be found here and a mocked up version for a fictional study is here. Detailed information on how to submit to the EVA can be found here.
Can my submitted data be held privately at the EVA?
Yes. Data submitted to the EVA can be held privately for up to one year. The date of publication is set by the submitter using the "Hold Date" field of the EVA metadata template.
What are the key steps for the dbSNP - EVA transitional process?
What metadata accessions are administered by the EVA?
The EVA follows the SRA model for accessioning metadata:
- A submitted project is administered a 'project' accession. These begin "PRJEB" followed by a numerical sequence.
- Each analysis object within a project is administered an 'analysis' accession. These begin "ERZ" followed by a numerical sequence.
Both the project and analysis accessions are sent to the submitter once the data has been validated and fully archived at EMBL-EBI. The project and analysis accessions are stable identifiers that are suitable for publication.
What variant accessions will be administered by the EVA?
The EVA shall follow the same accessioning principles employed by dbSNP: non-human variants that are submitted to the EVA are issued 'Submitted SNP' (SS) identifiers, and these are periodically clustered to form 'Reference SNP' (RS) identifiers.
Submitted SNP (SS)
A Submitted SNP (SS) identifies a variant submitted in the context of a genomic study. New SS IDs issued by the EVA will always be linked to a change from a reference allele to a single alternate allele. This means that a SS ID uniquely identifies a combination of:
- Taxonomy (e.g. 9615 - dog)
- Reference sequence (e.g. CanFam3.1)
- EVA study or dbSNP batch (e.g. PRJEB24066)
- Contig or chromosome
- Start position
- Reference allele
- Alternate allele
Most pre-existing dbSNP identifiers also follow this rule but, in some cases, multiple alternate alleles in the same location could be assigned a single SS ID depending on the representation at the time of submission.
An SS ID is linked to a variant in a particular study. Therefore, the same variant reported across multiple files in the study will still be assigned only one SS ID.
The following restrictions apply to EVA submissions, although legacy dbSNP identifiers are preserved even if they don’t:
- Variants must have a specific location in a sequence accessioned by INSDC in order to be archived and accessioned.
- Variants must have supporting evidence (genotypes and/or allele frequencies) in order to be archived and accessioned.
- Non-variants (as in gVCF reference blocks) will be archived, but will neither be assigned an SS ID nor displayed in the Variant Browser.
The list of SS IDs issued for a submission are provided to the submitter in VCF format, as in this example; please note the SS IDs listed there are mere examples and do not refer to real variants. Multi-allelic variants are split in multiple records, one per alternate allele. Variants are always reported in the forward strand.
Reference SNP (RS)
A Reference SNP (RS) identifies a variant class in a genomic location, which can be reported by one or more studies. RS IDs are preserved when remapped to a newer version of the reference sequence they were initially mapped against; this allows to track how they have evolved with the passing of time. This means that a RS ID can be associated with one or more combinations of:
- Taxonomy (e.g. 9615 - dog)
- Reference sequence (e.g. CanFam3.1)
- Contig or chromosome
- Start position
- Variant class
The following table lists the mapping between variant classes supported by dbSNP and the EVA, and the Sequence Ontology terms used by the latter.
dbSNP class EVA class Sequence Ontology term SNV SNV SO:0001483 DIV deletion SO:0000159 insertion SO:0000667 indel SO:1000032 HETEROZYGOUS STR / microsatellite tandem repeat SO:0000705 NAMED sequence alteration SO:0001059 NO VARIATION no sequence alteration SO:0002073 MIXED MNV MNV SO:0002007
Please note that due to the different granularities, dbSNP DIV variants will be assigned the appropriate EVA class when imported.
HETEROZYGOUS and MIXED variants were initially defined but no longer supported in any of the non-human dbSNP databases, thus why the EVA hasn’t defined a mapping.
Variants of class NO VARIATION will no longer be accessioned by the EVA, but legacy identifiers will be preserved.
The list of RS IDs issued for a species in a release will be provided in VCF format, as in this example; multi-allelic variants are split in multiple records, one per alternate allele, even if several of them have the same variant class. Coordinates are mandatory in VCF format, so RefSNP's without them won't be included in the dump. Please note the example file is the actual RS release for the barrel medic species (Medicago truncatula).
Will it be possible to retrieve existing dbSNP identifiers from the EVA?
Yes. The EVA is committed to the continuation of existing dbSNP identifiers for Submitted SNP (SS) and Reference SNP (RS).
The identifiers of those dbSNP variants that satisfy the EVA submission requirements will be retrievable via the EVA variant browser and web services API. If you want to learn more about EVA submission requirements, please click here.
dbSNP variants that don't satisfy these requirements will still be imported and searchable via a specific API and web view. We will work to make this experience as intuitive as possible, while keeping our commitment to only make high-quality variants part of the core EVA database.
Why are some variants in a different strand than the one dbSNP reports?
The EVA accepts variant submission in the forward strand only, whereas dbSNP is not so restrictive. In addition to this, dbSNP registers multiple orientations:
- Contig to chromosome
- SNP to contig
- SubSNP (ss) to RefSNP (rs)
As a result, variants with a 'reverse' orientation will be displayed differently despite the underlying data being equivalent.
Why can't I find some of the dbSNP identifiers on the EVA website?
We will gradually import the dbSNP data, and your species of interest may not have been yet processed. Please check our status report on the transition process.
If a species has been fully imported and your variant of interest is not displayed in our browser, please communicate the issue via email@example.com
Will there be EVA builds or releases like the dbSNP ones?
Yes. The EVA will create RefSNP dumps in VCF format every 6 months, starting in Q3 2018. These dumps shall contain the most basic information about each RefSNP:
- Genomic coordinates
- Reference and alternate alleles in the forward strand
- Identifiers of the studies that reported the variant
The list of RS IDs issued for a species in a release will be provided in VCF format, as in this example; multi-allelic variants are split in multiple records, one per alternate allele, even if several of them have the same variant class. Coordinates are mandatory in VCF format, so RefSNP's without them won't be included in the dump. Please note the RS IDs listed in the file are mere examples and do not refer to real variants.
The details of a study submission, reported genotypes and frequencies, etc., can then be queried using the existing EVA website and web services API.
EVA & dbSNP Transition
EMBL-EBI will maintain reliable accessions for non-human genetic variation data through the European Variation Archive (EVA). NCBI's dbSNP database will continue to maintain stable identifiers for human genetic variation data only. This change will enable a more rapid turnaround for data sharing in this burgeoning field.
For more information, please subscribe to firstname.lastname@example.org and follow us at @evarchive, where we shall continue to post updates. If anything remains unclear please contact us at email@example.com.