gnomAD

Overview

The Genome Aggregation Database (gnomAD) is a resource developed by an international coalition of investigators, with the goal of aggregating and harmonizing both exome and genome sequencing data from a wide variety of large-scale sequencing projects, and making summary data available for the wider scientific community.

Publication

Koch, L., 2020. Exploring human genomic diversity with gnomAD. Nature Reviews Genetics, 21(8), pp.448-448.

Illumina Connected Analysis will support gnomAD v4.0 for GRCh38 assembly and gnomAD v2.1 for GRCh37.

gnomAD v4.0 (GRCh38)

Small Variants

In gnomAD v4.0, like gnomAD v2.1, there are genome and exome data. Compare to gnomAD v2.1 which the data for genome and exome are merged, for gnomAD 4.0, Illumina Connected Annotation will separate them with different JSON output field. For gnomAD genome, the field name would be gnomad. For gnomAD exome, the field name would be gnomad-exome. Despite this difference in the field name, the JSON data format would be identical for both genome and exome.

VCF extraction

We currently extract the following info fields from both gnomAD genome and exome VCF files:

##INFO=<ID=AC,Number=A,Type=Integer,Description="Alternate allele count">
##INFO=<ID=AN,Number=1,Type=Integer,Description="Total number of alleles">
##INFO=<ID=AC_XX,Number=A,Type=Integer,Description="Alternate allele count for XX samples">
##INFO=<ID=AN_XX,Number=1,Type=Integer,Description="Total number of alleles in XX samples">
##INFO=<ID=nhomalt_XX,Number=A,Type=Integer,Description="Count of homozygous individuals in XX samples">
##INFO=<ID=AC_XY,Number=A,Type=Integer,Description="Alternate allele count for XY samples">
##INFO=<ID=AN_XY,Number=1,Type=Integer,Description="Total number of alleles in XY samples">
##INFO=<ID=nhomalt_XY,Number=A,Type=Integer,Description="Count of homozygous individuals in XY samples">
##INFO=<ID=nhomalt,Number=A,Type=Integer,Description="Count of homozygous individuals">
##INFO=<ID=AC_afr,Number=A,Type=Integer,Description="Alternate allele count for samples of African/African-American ancestry">
##INFO=<ID=AN_afr,Number=1,Type=Integer,Description="Total number of alleles in samples of African/African-American ancestry">
##INFO=<ID=nhomalt_afr,Number=A,Type=Integer,Description="Count of homozygous individuals in samples of African/African-American ancestry">
##INFO=<ID=AC_amr,Number=A,Type=Integer,Description="Alternate allele count for samples of Latino ancestry">
##INFO=<ID=AN_amr,Number=1,Type=Integer,Description="Total number of alleles in samples of Latino ancestry">
##INFO=<ID=nhomalt_amr,Number=A,Type=Integer,Description="Count of homozygous individuals in samples of Latino ancestry">
##INFO=<ID=AC_asj,Number=A,Type=Integer,Description="Alternate allele count for samples of Ashkenazi Jewish ancestry">
##INFO=<ID=AN_asj,Number=1,Type=Integer,Description="Total number of alleles in samples of Ashkenazi Jewish ancestry">
##INFO=<ID=nhomalt_asj,Number=A,Type=Integer,Description="Count of homozygous individuals in samples of Ashkenazi Jewish ancestry">
##INFO=<ID=AC_eas,Number=A,Type=Integer,Description="Alternate allele count for samples of East Asian ancestry">
##INFO=<ID=AN_eas,Number=1,Type=Integer,Description="Total number of alleles in samples of East Asian ancestry">
##INFO=<ID=nhomalt_eas,Number=A,Type=Integer,Description="Count of homozygous individuals in samples of East Asian ancestry">
##INFO=<ID=AC_fin,Number=A,Type=Integer,Description="Alternate allele count for samples of Finnish ancestry">
##INFO=<ID=AN_fin,Number=1,Type=Integer,Description="Total number of alleles in samples of Finnish ancestry">
##INFO=<ID=nhomalt_fin,Number=A,Type=Integer,Description="Count of homozygous individuals in samples of Finnish ancestry">
##INFO=<ID=AC_mid,Number=A,Type=Integer,Description="Alternate allele count for samples of Middle Eastern ancestry">
##INFO=<ID=AN_mid,Number=1,Type=Integer,Description="Total number of alleles in samples of Middle Eastern ancestry">
##INFO=<ID=nhomalt_mid,Number=A,Type=Integer,Description="Count of homozygous individuals in samples of Middle Eastern ancestry">
##INFO=<ID=AC_nfe,Number=A,Type=Integer,Description="Alternate allele count for samples of Non-Finnish European ancestry">
##INFO=<ID=AN_nfe,Number=1,Type=Integer,Description="Total number of alleles in samples of Non-Finnish European ancestry">
##INFO=<ID=nhomalt_nfe,Number=A,Type=Integer,Description="Count of homozygous individuals in samples of Non-Finnish European ancestry">
##INFO=<ID=AC_remaining,Number=A,Type=Integer,Description="Alternate allele count for samples of Remaining individuals ancestry">
##INFO=<ID=AN_remaining,Number=1,Type=Integer,Description="Total number of alleles in samples of Remaining individuals ancestry">
##INFO=<ID=nhomalt_remaining,Number=A,Type=Integer,Description="Count of homozygous individuals in samples of Remaining individuals ancestry">
##INFO=<ID=AC_sas,Number=A,Type=Integer,Description="Alternate allele count for samples of South Asian ancestry">
##INFO=<ID=AN_sas,Number=1,Type=Integer,Description="Total number of alleles in samples of South Asian ancestry">

JSON output

Field	Type	Notes
coverage	int	average coverage (non-negative integer values)
maleAf	float	allele frequency for male population. Range: 0 - 1.0
maleAn	int	allele number for male population. Non-zero integer.
maleAc	int	allele count for male population. Integer.
maleHc	int	count of homozygous individuals for male population. Non-negative integer.
femaleAf	float	allele frequency for female population. Range: 0 - 1.0
femaleAn	int	allele number for female population. Non-zero integer.
femaleAc	int	allele count for female population. Integer.
femaleHc	int	count of homozygous individuals for female population. Non-negative integer.
remainingAf	float	allele frequency for the Other population. Range: 0 - 1.0
remainingAc	int	allele count for the Other population. Integer.
remainingAn	int	allele number for the Other population. Non-zero integer.
remainingHc	int	count of homozygous individuals for Other population. Non-negative integer
allAf	float	allele frequency for all populations. Range: 0 - 1.0
allAn	int	allele number for all populations. Non-zero integer.
allAc	int	allele count for all populations. Integer.
allHc	int	count of homozygous individuals for all populations. Non-negative integer.
afrAf	float	allele frequency for the African / African American population. Range: 0 - 1.0
afrAc	int	allele count for the African / African American population. Integer.
afrAn	int	allele number for the African / African American population. Non-zero integer.
afrHc	int	count of homozygous individuals for African / African American population. Non-negative integer.
amiAf	float	allele frequency for Amish populations. Range: 0 - 1.0
amiAn	int	allele number for Amish populations. Non-zero integer.
amiAc	int	allele count for Amish populations. Integer.
amiHc	int	count of homozygous individuals for Amish populations. Non-negative integer.
amrAf	float	allele frequency for the Latino population. Range: 0 - 1.0
amrAc	int	allele count for the Latino population. Integer.
amrAn	int	allele number for the Latino population. Non-zero integer.
amrHc	int	count of homozygous individuals for Latino population. Non-negative integer.
easAf	float	allele frequency for the East Asian population. Range: 0 - 1.0
easAc	int	allele count for the East Asian population. Integer.
easAn	int	allele number for the East Asian population. Non-zero integer.
easHc	int	count of homozygous individuals for East Asian population. Non-negative integer.
finAf	float	allele frequency for the Finnish population. Range: 0 - 1.0
finAc	int	allele count for the Finnish population. Integer.
finAn	int	allele number for the Finnish population. Non-zero integer.
finHc	int	count of homozygous individuals for Finnish population. Non-negative integer
nfeAf	float	allele frequency for the Non-Finnish European population. Range: 0 - 1.0
nfeAc	int	allele count for the Non-Finnish European population. Integer.
nfeAn	int	allele number for the Non-Finnish European population. Non-zero integer.
nfeHc	int	count of homozygous individuals for Non-Finnish European population. Non-negative integer
asjAf	float	allele frequency for the Ashkenazi Jewish population. Range: 0 - 1.0
asjAc	int	allele count for the Ashkenazi Jewish population Integer.
asjAn	int	allele number for the Ashkenazi Jewish population. Non-zero integer.
asjHc	int	count of homozygous individuals for the Ashkenazi Jewish population. Non-negative integer
sasAf	float	allele frequency for the South Asian population. Range: 0 - 1.0
sasAc	int	allele count for the South Asian population Integer.
sasAn	int	allele number for the South Asian population. Non-zero integer.
sasHc	int	count of homozygous individuals for the South Asian population. Non-negative integer.
midAf	float	allele frequency for the Middle Eastern population. Range: 0 - 1.0
midAc	int	allele count for the iddle Eastern population Integer.
midAn	int	allele number for the iddle Eastern population. Non-zero integer.
midHc	int	count of homozygous individuals for the iddle Eastern population. Non-negative integer.
failedFilter	bool	True if this variant failed any filters (Note: we do not list the failed filters)

Calculation

To calculate allele frequency for each group, we divide the allele count with allele number for each group.

LoF Gene Metrics

In gnomAD 4.0, the gene score data for LOF is given per transcript. Since this is gene level data, one of the transcripts need to be chosen and value reported. The transcript ID of the selected transcript will be reported. Transcripts are prioritized (from higher to lower) as follows:

1. Ensembl Transcript has mane_select column true from source (gnomAD).
2. Transcript is marked as Ensembl canonical in Illumina Connected Annotation cache data.
3. RefSeq transcript has mane_select column true.
4. Transcript is marked as RefSeq canonical in Illumina Connected Annotation cache data.
5. Transcript has the lowest lof.oe_ci.upper value compare to other transcript for the same gene.

Differences with gnomAD browser

Due to difference in Ensembl version between Illumina Connected Annotation and gnomAD, there are several transcript ID that are marked as canonical in gnomAD browser but not in Illumina Connected Analysis. If this is the case, the gene score shown in Illumina Connected Annotation will be different compared to the gene score shown in the gnomAD browser. The transcriptId field in the JSON output will report which transcript was used by Illumina Connected Annotation.

Tab delimited file example

JSON key to TSV column mapping

JSON key	TSV column	Description
pLi	lof.pLI	probability of being intolerant of a single loss-of-function variant (like haploinsufficient genes, observed ~ 0.1*expected)
pNull	lof.pNull	probability of being completely tolerant of loss of function variation (observed = expected)
pRec	lof.pRec	probability of being intolerant of two loss of function variants (like recessive genes, observed ~ 0.5*expected)
synZ	syn.z_score	corrected synonymous Z score
misZ	mis.z_score	corrected missense Z score
loeuf	lof.oe_ci.upper	loss of function observed/expected upper bound fraction (LOEUF)
transcriptId	transcript	transcript ID which the values we select

Structural Variants

Structural variants in gnomAD 4.0 is available in VCF format and has the same population data as small variants.

Structural Variant Type Mapping

The source files represented the structural variants with keys using various naming conventions. In the Illumina Connected Annotations JSON output, these keys will be mapped according to the following.

Illumina Connected Annotations JSON SV Type Key	GRCh37 Source SV Type Key
deletion	DEL, CN=0
duplication	DUP
insertion	INS
inversion	INV
mobile_element_insertion	INS:ME
mobile_element_insertion	INS:ME:ALU
mobile_element_insertion	INS:ME:LINE1
mobile_element_insertion	INS:ME:SVA
complex_structural_alteration	CPX

gnomAD Copy Number Variation

In gnomAD 4.0 structural variants data, there are CNV data in the VCF file. Since it is not shown in the browser, we don't include CNV in our output. We will evaluate in the future whether to include copy number variation from structural variation data together with new rare CNV data taht is available in gnomAD 4.0.

gnomAD duplication variant type

In gnomAD 4.0 structural variants VCF, only DUP is shown as symbolic allele for duplication variant type. Based on the information in gnomAD browser, duplication variant that has split read or paired end reads evidence can be inferred as tandem duplication. With this, we check the evidence data in each DUP variants entry to decide whether it can be assign tandem duplication as variant type or it is just duplication.

JSON output

gnomAD v2.1 (GRCh37)

Small Variants

VCF extraction

We currently extract the following info fields from gnomAD genome and exome VCF files:

We also extract the following extra fields from gnomAD exome VCF file:

Computation

Using these, we compute the following:

• Coverage
• Allele count, Homozygous count, allele number and allele frequencies for:
• Global population
• African/African Americans
• Admixed Americans
• Ashkenazi Jews
• East Asians
• Finnish
• Non-Finnish Europeans
• South Asian
• Others (population not assigned)
• Male
• Female
• Controls

Note

• Coverage = DP / AN. Frequencies are computed using AC/AN for each population.
• Please note that currently there is no genome sequencing data of south asian (SAS) population available in gnomAD.
• Allele Count, Homozygous count, allele number and allele frequencies for control groups are also provided for the global population.

Merging genomes and exomes

When merging the genomes and exomes, the allele counts and allele numbers will be summed across both of the data sets.

• For GRCh37, Illumina Connected Annotations currently uses gnomAD version 2.1 which contains both genomes and exomes data. Genomes and exomes data are merged in the output.

Filters

The following strategy will be used when there's a conflict in filter status:

	Genomes PASS	Genomes Filtered
Exomes PASS	PASS	Only use exome data
Exomes Filtered	Only use genome data	Filtered

VCF download instructions

https://gnomad.broadinstitute.org/downloads

JSON output

Field	Type	Notes
coverage	int	average coverage (non-negative integer values)
allAf	float	allele frequency for all populations. Range: 0 - 1.0
maleAf	float	allele frequency for male population. Range: 0 - 1.0
femaleAf	float	allele frequency for female population. Range: 0 - 1.0
controlsAllAf	float	allele frequency for the controls subset. Range: 0 - 1.0
allAc	int	allele count for all populations. Integer.
maleAc	int	allele count for male population. Integer.
femaleAc	int	allele count for female population. Integer.
controlsAllAc	int	allele count for the controls subset. Integer.
allAn	int	allele number for all populations. Non-zero integer.
maleAn	int	allele number for male population. Non-zero integer.
femaleAn	int	allele number for female population. Non-zero integer.
controlsAllAn	int	allele number for the controls subset. Non-zero integer.
allHc	int	count of homozygous individuals for all populations. Non-negative integer.
maleHc	int	count of homozygous individuals for male population. Non-negative integer.
femaleHc	int	count of homozygous individuals for female population. Non-negative integer.
afrAf	float	allele frequency for the African / African American population. Range: 0 - 1.0
afrAc	int	allele count for the African / African American population. Integer.
afrAn	int	allele number for the African / African American population. Non-zero integer.
afrHc	int	count of homozygous individuals for African / African American population. Non-negative integer.
amrAf	float	allele frequency for the Latino population. Range: 0 - 1.0
amrAc	int	allele count for the Latino population. Integer.
amrAn	int	allele number for the Latino population. Non-zero integer.
amrHc	int	count of homozygous individuals for Latino population. Non-negative integer.
easAf	float	allele frequency for the East Asian population. Range: 0 - 1.0
easAc	int	allele count for the East Asian population. Integer.
easAn	int	allele number for the East Asian population. Non-zero integer.
easHc	int	count of homozygous individuals for East Asian population. Non-negative integer.
finAf	float	allele frequency for the Finnish population. Range: 0 - 1.0
finAc	int	allele count for the Finnish population. Integer.
finAn	int	allele number for the Finnish population. Non-zero integer.
finHc	int	count of homozygous individuals for Finnish population. Non-negative integer
nfeAf	float	allele frequency for the Non-Finnish European population. Range: 0 - 1.0
nfeAc	int	allele count for the Non-Finnish European population. Integer.
nfeAn	int	allele number for the Non-Finnish European population. Non-zero integer.
nfeHc	int	count of homozygous individuals for Non-Finnish European population. Non-negative integer
othAf	float	allele frequency for the Other population. Range: 0 - 1.0
othAc	int	allele count for the Other population. Integer.
othAn	int	allele number for the Other population. Non-zero integer.
othHc	int	count of homozygous individuals for Other population. Non-negative integer
asjAf	float	allele frequency for the Ashkenazi Jewish population. Range: 0 - 1.0
asjAc	int	allele count for the Ashkenazi Jewish population Integer.
asjAn	int	allele number for the Ashkenazi Jewish population. Non-zero integer.
asjHc	int	count of homozygous individuals for the Ashkenazi Jewish population. Non-negative integer
sasAf	float	allele frequency for the South Asian population. Range: 0 - 1.0
sasAc	int	allele count for the South Asian population Integer.
sasAn	int	allele number for the South Asian population. Non-zero integer.
sasHc	int	count of homozygous individuals for the South Asian population. Non-negative integer.
failedFilter	bool	True if this variant failed any filters (Note: we do not list the failed filters)
lowComplexityRegion	bool	True if this variant is located in a low complexity region.

Building the supplementary files

The gnomAD .nsa for Illumina Connected Annotations can be built using the SAUtils command's gnomad subcommand. We will describe building gnomAD version 3.1 here.

Source data files

Input VCF files (one per chromosome) and a .version file are required in a folder to build the .nsa file. For example, my directory contains:

The version file is a text file with the following content.

The help menu for the utility is as follows:

Here is a sample execution:

LoF Gene Metrics

Tab delimited file example

JSON key to TSV column mapping

JSON key	TSV column	Description
pLi	pLI	probability of being intolerant of a single loss-of-function variant (like haploinsufficient genes, observed ~ 0.1*expected)
pNull	pNull	probability of being completely tolerant of loss of function variation (observed = expected)
pRec	pRec	probability of being intolerant of two loss of function variants (like recessive genes, observed ~ 0.5*expected)
synZ	syn_z	corrected synonymous Z score
misZ	mis_z	corrected missense Z score
loeuf	oe_lof_upper	loss of function observed/expected upper bound fraction (LOEUF)

Gene symbol update

The input file provides Ensembl gene ids for each entry. We observed that they were unique while gene symbols may be repeated (multiple lines may have the same gene symbol). Since Ensembl gene Ids are more stable, and Illumina Connected Annotations transcript cache data contains Ensembl gene ids, we use these ids to extract the gene symbols from the transcript cache. For example, if ENSG0001 has gene symbol GENE1 in the input but Illumina Connected Annotations cache say ENSG0001 maps to GENE2, we use GENE2 as the gene symbol for that entry.

Conflict resolution

gnomAD uses Ensembl GeneID as unique identifiers in the source file but Illumina Connected Annotations uses HGNC gene symbols. Multiple Ensembl GeneIDs can map to the same HGNC symbol and therefore may result is conflict.

In such cases, Illumina Connected Annotations chooses the entry with the smallest "LOEUF" value. The reason for choosing this value can be highlighted by the following table:

LOEUF decile	Haplo-insufficient	Autosomal Dominant	Autosomal Recessive	Olfactory Genes
0-10%	104	140	36	0
10-20%	47	128	72	1
20-30%	17	86	112	0
30-40%	8	80	173	4
40-50%	7	65	206	8
50-60%	4	54	207	6
60-70%	0	46	154	18
70-80%	2	49	120	49
80-90%	0	34	58	96
90-100%	0	26	40	174

Note

• Table source: https://www.biorxiv.org/content/biorxiv/early/2019/01/28/531210.full-text.pdf
• This table indicates that lower LOEUF scores have more deleterious effect on genes.
• Only 15 out of 19685 genes have conflicting entries.

List of genes with conflicting entries

Conflict resolution

• Pick the entry with the lowest LOEUF score
• If the same, pick the lowest pLI
• Otherwise pick the entry with the max absolute value of synZ + misZ

Download URL

https://storage.googleapis.com/gnomad-public/release/2.1.1/constraint/gnomad.v2.1.1.lof_metrics.by_gene.txt.bgz

JSON output

Field	Type	Notes
pLi	float	probability of being intolerant of a single loss-of-function variant (like haploinsufficient genes, observed ~ 0.1*expected)
pNull	float	probability of being completely tolerant of loss of function variation (observed = expected)
pRec	float	probability of being intolerant of two loss of function variants (like recessive genes, observed ~ 0.5*expected)
synZ	float	corrected synonymous Z score
misZ	float	corrected missense Z score
loeuf	float	loss of function observed/expected upper bound fraction (LOEUF)

Structural Variants

Publication

Collins, R.L., Brand, H., Karczewski, K.J. et al. 2020. A structural variation reference for medical and population genetics. Nature 581, pp.444–451. https://doi.org/10.1038/s41586-020-2287-8

Note The gnomAD structural variant annotations are in a preview stage at the moment. Currently, the annotations do not include translocation breakends. Future updates will include a better way of annotating the structural variants.

Source Files

Bed Example

The bed file was obtained from original source for GRCh37

Structural Variant Type Mapping

The source files represented the structural variants with keys using various naming conventions. In the Illumina Connected Annotations JSON output, these keys will be mapped according to the following.

Illumina Connected Annotations JSON SV Type Key	GRCh37 Source SV Type Key
copy_number_variation
deletion	DEL, CN=0
duplication	DUP
insertion	INS
inversion	INV
mobile_element_insertion	INS:ME
mobile_element_insertion	INS:ME:ALU
mobile_element_insertion	INS:ME:LINE1
mobile_element_insertion	INS:ME:SVA
structural alteration
complex_structural_alteration	CPX

Download URLs

GRCh37

The GRCh37 file was downloaded from the original source. Following table gives some essential data metrics:

https://storage.googleapis.com/gcp-public-data--gnomad/papers/2019-sv/gnomad_v2.1_sv.sites.bed.gz

JSON output

Field	Type	Notes
chromosome	string	chromosome number
begin	integer	position interval start
end	integer	position internal end
variantType	string	structural variant type
variantId	string	gnomAD ID
allAf	floating point	allele frequency for all populations. Range: 0 - 1.0
afrAf	floating point	allele frequency for the African super population. Range: 0 - 1.0
amrAf	floating point	allele frequency for the Ad Mixed American super population. Range: 0 - 1.0
easAf	floating point	allele frequency for the East Asian super population. Range: 0 - 1.0
eurAf	floating point	allele frequency for the European super population. Range: 0 - 1.0
othAf	floating point	allele frequency for all other populations. Range: 0 - 1.0
femaleAf	floating point	allele frequency for female population. Range: 0 - 1.0
maleAf	floating point	allele frequency for male population. Range: 0 - 1.0
allAc	integer	allele count for all populations.
afrAc	integer	allele count for the African super population.
amrAc	integer	allele count for the Ad Mixed American super population.
easAc	integer	allele count for the East Asian super population.
eurAc	integer	allele count for the European super population.
othAc	integer	allele count for all other populations.
maleAc	integer	allele count for male population.
femaleAc	integer	allele count for female population.
allAn	integer	allele number for all populations.
afrAn	integer	allele number for the African super population.
amrAn