Description

With the gnomAD v4.1 data release, the v4 Pre-Release track has been replaced with the gnomAD v4.1 track. The v4.1 release includes a fix for the allele number issue. The v4.1 track shows variants from 807,162 individuals, including 730,947 exomes and 76,215 genomes. This includes the 76,156 genomes from the gnomAD v3.1.2 release as well as new exome data from 416,555 UK Biobank individuals. For more detailed information on gnomAD v4.1, see the related blog post.

The gnomAD v3.1 track shows variants from 76,156 whole genomes (and no exomes), all mapped to the GRCh38/hg38 reference sequence. 4,454 genomes were added to the number of genomes in the previous v3 release. For more detailed information on gnomAD v3.1, see the related blog post.

The gnomAD v3.1.1 track contains the same underlying data as v3.1, but with minor corrections to the VEP annotations and dbSNP rsIDs. On the UCSC side, we have now included the mitochondrial chromosome data that was released as part of gnomAD v3.1 (but after the UCSC version of the track was released). For more information about gnomAD v3.1.1, please see the related changelog.

GnomAD Genome Mutational Constraint is based on v3.1.2 and is available only on hg38. It shows the reduced variation caused by purifying natural selection. This is similar to negative selection on loss-of-function (LoF) for genes, but can be calculated for non-coding regions too. Positive values are red and reflect stronger mutation constraint (and less variation), indicating higher natural selection pressure in a region. Negative values are green and reflect lower mutation constraint (and more variation), indicating less selection pressure and less functional effect. Briefly, for any 1kbp window in the genome, a model based on trinucleotide sequence context, base-level methylation, and regional genomic features predicts expected number of mutations, and compares this number to the observed number of mutations using a Z-score (see preprint in the Reference section for details). The chrX scores were added as received from the authors, as there are no de novo mutation data available on chrX (for estimating the effects of regional genomic features on mutation rates), they are more speculative than the ones on the autosomes.

The gnomAD Predicted Constraint Metrics track contains metrics of pathogenicity per-gene as predicted for gnomAD v2.1.1 and identifies genes subject to strong selection against various classes of mutation. This includes data on both the gene and transcript level.

The gnomAD v2 tracks show variants from 125,748 exomes and 15,708 whole genomes, all mapped to the GRCh37/hg19 reference sequence and lifted to the GRCh38/hg38 assembly. The data originate from 141,456 unrelated individuals sequenced as part of various population-genetic and disease-specific studies collected by the Genome Aggregation Database (gnomAD), release 2.1.1. Raw data from all studies have been reprocessed through a unified pipeline and jointly variant-called to increase consistency across projects. For more information on the processing pipeline and population annotations, see the following blog post and the 2.1.1 README.

gnomAD v2 data are based on the GRCh37/hg19 assembly. These tracks display the GRCh38/hg38 lift-over provided by gnomAD on their downloads site.

On hg38 only, a subtrack "Gnomad mutational constraint" aka "Genome non-coding constraint of haploinsufficient variation (Gnocchi)" captures the depletion of variation caused by purifying natural selection. This is similar to negative selection on loss-of-function (LoF) for genes, but can be calculated for non-coding regions, too. Briefly, for any 1kbp window in the genome, a model based on trinucleotide sequence context, base-level methylation, and regional genomic features predicts expected number of mutations, and compares this number to the observed number of mutations using a Z-score (see Chen et al 2024 in the Reference section for details). The chrX scores were added as received from the authors, as there are no mutations available for chrX, they are more speculative than the ones on the autosomes.

For questions on the gnomAD data, also see the gnomAD FAQ.

More details on the Variant type(s) can be found on the Sequence Ontology page.

Display Conventions and Configuration

gnomAD v4.1

The gnomAD v4.1 track version follows the same conventions and configuration as the v3.1.1 track, except for mouse hovering items. Mouse hover on an item will display the following details about each variant:

• Position
• Total Allele Frequency (TotalAF)
• Genes
• Annotation
• FILTER tags from VCF (FILTER)
• Population with maximum AF (PopMaxAF)
• Homozygous Individuals
• Homozygous Individuals in XX samples (chrX and chrY only)
• Hemizygous Individuals (chrX and chrY only)

gnomAD v3.1.1

The gnomAD v3.1.1 track version follows the same conventions and configuration as the v3.1 track, except as noted below.

1. There are additional FILTER field filters: AS_VQSR, indel_stack (chrM only), and npg (chrM only).
2. Where possible, variants overlapping multiple transcripts/genes have been collapsed into one variant, with additional information available on the details page, which has roughly halved the number of items in the bigBed.
3. The bigBed has been split into two files, one with the information necessary for the track display, and one with the information necessary for the details page. For more information on this data format, please see the Data Access section below.
4. The VEP annotation is shown as a table instead of spread across multiple fields.
5. Intergenic variants have not been pre-filtered.

gnomAD v3.1

By default, a maximum of 50,000 variants can be displayed at a time (before applying the filters described below), before the track switches to dense display mode.

Mouse hover on an item will display many details about each variant, including the affected gene(s), the variant type, and annotation (missense, synonymous, etc).

Clicking on an item will display additional details on the variant, including a population frequency table showing allele count in each sub-population.

Following the conventions on the gnomAD browser, items are shaded according to their Annotation type:

Label Options

To maintain consistency with the gnomAD website, variants are by default labeled according to their chromosomal start position followed by the reference and alternate alleles, for example "chr1-1234-T-CAG". dbSNP rsID's are also available as an additional label, if the variant is present in dbSnp.

Filtering Options

Three filters are available for these tracks:

• FILTER: Used to exclude/include variants that failed Random Forest (RF), Inbreeding Coefficient (Inbreeding Coeff), or Allele Count (AC0) filters. The PASS option is used to include/exclude variants that pass all of the RF, InbreedingCoeff, and AC0 filters, as denoted in the original VCF.
• Annotation type: Used to exclude/include variants that are annotated as Probability Loss of Function (pLoF), Missense, Synonymous, or Other, as annotated by VEP version 85 (GENCODE v19).
• Variant Type: Used to exclude/include variants according to the type of variation, as annotated by VEP v85.

gnomAD v2.1.1

The gnomAD v2.1.1 track follows the standard display and configuration options available for VCF tracks, briefly explained below.

• In dense mode, a vertical line is drawn at the position of each variant.
• In pack mode, "ref" and "alt" alleles are displayed to the left of a vertical line with colored portions corresponding to allele counts. Hovering the mouse pointer over a variant pops up a display of alleles and counts.

Filtering Options

Four filters are available for these tracks, the same as the underlying VCF:

• AC0: Allele Count 0 after filtering out low confidence genotypes (GQ < 20; DP < 10; and AB < 0.2 for het calls))
• InbreedingCoeff: Inbreeding Coefficient < -0.3
• RF: Used to exclude/include variants that failed Random Forest filtering thresholds of 0.055272738028512555, 0.20641025579497013 (probabilities of being a true positive variant) for SNPs, indels)
• Pass: Variant passes all 3 filters

There are two additional filters available, one for the minimum minor allele frequency, and a configurable filter on the QUAL score.

For the full steps used to create the gnomAD tracks at UCSC, please see the hg38 gnomad makedoc.

The raw data can be explored interactively with the Table Browser, or the Data Integrator. For automated analysis, the data may be queried from our REST API, and the genome annotations are stored in files that can be downloaded from our download server, subject to the conditions set forth by the gnomAD consortium (see below). Variant VCFs can be found in the vcf subdirectory. The v3.1, v3.1.1, and v4.1 variants can be found in a special directory as they have been transformed from the underlying VCF.

For the v3.1.1 and v4.1 variants in particular, the underlying bigBed only contains enough information necessary to use the track in the browser. The extra data like VEP annotations and CADD scores are available in the same directory as the bigBed but in the files details.tab.gz and details.tab.gz.gzi. The details.tab.gz contains the gzip compressed extra data in JSON format, and the .gzi file is available to speed searching of this data. Each variant has an associated md5sum in the name field of the bigBed which can be used along with the _dataOffset and _dataLen fields to get the associated external data, as show below:

# find item of interest:
bigBedToBed genomes.bb stdout | head -4 | tail -1
chr1    12416    12417    854246d79dc5d02dcdbd5f5438542b6e    [..omitted for brevity..]    chr1-12417-G-A    67293    902

# use the final two fields, _dataOffset and _dataLen (add one to _dataLen to include a newline), to get the extra data:
bgzip -b 67293 -s 903 gnomad.v3.1.1.details.tab.gz
854246d79dc5d02dcdbd5f5438542b6e    {"DDX11L1": {"cons": ["non_coding_transcript_variant",  [..omitted for brevity..]

The data can also be found directly from the gnomAD downloads page. Please refer to our mailing list archives for questions, or our Data Access FAQ for more information.

The mutational constraints score was updated in October 2022 from a previous, now deprecated, pre-publication version. The old version can be found in our archive directory on the download server. It can be loaded by copying the URL into our "Custom tracks" input box.

Credits

Thanks to the Genome Aggregation Database Consortium for making these data available. The data are released under the Creative Commons Zero Public Domain Dedication as described here.

Please note that some annotations within the provided files may have restrictions on usage. See here for more information.

References

Karczewski KJ, Francioli LC, Tiao G, Cummings BB, Alfoldi J, Wang Q, Collins RL, Laricchia KM, Ganna A, Birnbaum DP et al. Variation across 141,456 human exomes and genomes reveals the spectrum of loss-of-function intolerance across human protein-coding genes. doi: https://doi.org/10.1101/531210.

Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, O'Donnell-Luria AH, Ware JS, Hill AJ, Cummings BB et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature. 2016 Aug 17;536(7616):285-91. PMID: 27535533; PMC: PMC5018207

Chen S, Francioli LC, Goodrich JK, Collins RL, Kanai M, Wang Q, Alföldi J, Watts NA, Vittal C, Gauthier LD et al. A genomic mutational constraint map using variation in 76,156 human genomes. Nature. 2024 Jan;625(7993):92-100. PMID: 38057664
(We added the data in 2021, then later referenced the 2022 Biorxiv preprint, in which the track was not called "Gnocchi" yet)