UK Biobank Pharma Proteomics Project

Data Source

ID: ukb_ppp

URL: https://doi.org/10.7303/syn51364943

License: CC BY

Citation:

Sun BB, Chiou J, Traylor M, et al. Plasma proteomic associations with genetics and health in the UK Biobank. Nature. 622, 329-338. (2023).

Datasets

• ukb_ppp.assays
• ukb_ppp.genes
• ukb_ppp.pqtls
• ukb_ppp.proteins
• ukb_ppp.variants

Ancestries

• AFR = African
• AMR = American
• ALL = All ancestries
• CSA = Central-South Asian
• EAS = East Asian
• EUR = European
• MID = Middle Eastern

Harmonization

Note

• GRCh37 and GRCh38 coordinates were reported in source files. No liftover was necessary.
• The dbsnp.vcf table (build == 'b157', assembly == 'GRCh38') was used as the reference forward strand.
• Study variants were aligned to the reference forward strand, so that effect_allele == alt_allele and other_allele == ref_allele.
• Strand consensus approach was used to harmonize palindromic variants: consensus strand was inferred from non-palindromic variants, and that strand was assumed for all palindromic variants. This is the approach used by the Open Targets Mendelian randomization pipeline.

Caution

In the ukb_ppp.variants table, a small (<<1%) number of variants have reported_strand == 'reverse' (in relation to the reference forward strand). For non-palindromic variants, these variants can be identified and transformed appropriately.

A logical assumption would be that an equal proportion of palindromic variants would also have reported_strand == 'reverse', but these cannot be definitively identified.

However, with the strand consensus approach, we assume the same strand for all palindromic variants, so there is likely a very small fraction of palindromic variants for which the reported_strand (and thus the applied transformation) is inaccurate.

The most conservative approach is to exclude palindromic variants from downstream analyses, or to further harmonize with an allele frequency-based approach. This decision is left to the user.

---

is_palindromic	in_dbsnp	reported_strand	was_flipped	Outcome
False	True	'forward'	False	Non-palindromic, effect_allele == alt_allele and other_allele == ref_allele -> no changes.
False	True	'forward'	True	Non-palindromic, effect_allele == ref_allele and other_allele == alt_allele -> swap alleles, flip beta and effect_allele_frequency.
False	True	'reverse'	False	Non-palindromic, reverse_complement(effect_allele) == alt_allele and reverse_complement(other_allele) == ref_allele -> take reverse complement of alleles, beta and effect_allele_frequency unchanged.
False	True	'reverse'	True	Non-palindromic, reverse_complement(effect_allele) == ref_allele and reverse_complement(other_allele) == alt_allele -> swap and take reverse complement of alleles, flip beta and effect_allele_frequency.
False	False	NULL	NULL	Non-palindromic, no variant match in reference -> unharmonized. Alleles, beta, and effect_allele_frequency unchanged. Proceed with caution.
True	True	'forward' (inferred from non-palindromic consensus)	False	Palindromic, effect_allele == alt_allele and other_allele == ref_allele -> no changes. Strand is inferred rather than known definitively, proceed with slight caution.
True	True	'forward' (inferred from non-palindromic consensus)	True	Palindromic, effect_allele == ref_allele and other_allele == alt_allele -> swap alleles, flip beta and effect_allele_frequency. Strand is inferred rather than known definitively, proceed with slight caution.
True	True	'reverse' (inferred from non-palindromic consensus)	False	Palindromic, reverse_complement(effect_allele) == alt_allele and reverse_complement(other_allele) == ref_allele -> take reverse complement of alleles, beta and effect_allele_frequency unchanged. Strand is inferred rather than known definitively, proceed with slight caution.
True	True	'reverse' (inferred from non-palindromic consensus)	True	Palindromic, reverse_complement(effect_allele) == ref_allele and reverse_complement(other_allele) == alt_allele -> swap and take reverse complement of alleles, flip beta and effect_allele_frequency. Strand is inferred rather than known definitively, proceed with slight caution.
True	False	NULL	NULL	Palindromic, no variant match in reference -> unharmonized. Alleles, beta, and effect_allele_frequency unchanged. Proceed with caution.

---

assays

Column	Type	Description
partition	INT	Dummy partition column to enable compatibility with Cloudflare R2 SQL API. Ignore.
ukb_ppp_id	TEXT	Unique assay identifier defined by UKB-PPP.
olink_id	TEXT	Olink platform assay identifier (e.g. OID30809).
gene_symbol	TEXT	HGNC gene symbol of the gene coding the target protein (e.g. GLP1R, APOE, TNF).
panel	TEXT	Olink panel containing the assay.
panel_lot	TEXT	Olink lot identifier of the panel.
dilution_factor	INT	Dilution factor of the assay.
block	TEXT	Identifier of the 96-well block within the 384-plex Olink panel.
in_expansion_set	BOOLEAN	Whether the assay was part of the original set (false) or added in the expansion set (true).
protein_id	TEXT	UniProt accession (e.g. P43220, P02649), or multiple accessions joined by underscore for multi-chain protein complexes.
filename	TEXT	Name of the tar archive file containing pQTL results for the assay.

genes

Column	Type	Description
partition	INT	Dummy partition column to enable compatibility with Cloudflare R2 SQL API. Ignore.
gene_id	TEXT	Ensembl gene identifier (e.g. ENSG00000112164).
gene_symbol	TEXT	HGNC gene symbol (e.g. GLP1R, APOE).
chromosome	TEXT	Chromosome on which the gene is located.
start_position	INT	Gene start position in assembly GRCh38 coordinates.
end_position	INT	Gene end position in assembly GRCh38 coordinates.
strand	TEXT	Strand of the gene.

pqtls

Note

All load_dataset() queries must filter on partition columns: ancestry, protein_id, panel.

Note

See ukb_ppp.assays dataset for gene_symbol, protein_id, panel mappings and values.

Note

Most proteins were measured in a single assay panel and will have a one-to-one mapping between protein_id and panel.

Six proteins were measured in multiple assay panels, and so will have a one-to-multiple mapping between protein_id and panel. This is the reason for including the panel as a partition column.

Note

pQTL associations are missing from the source for two proteins in the AMR population: Q15113 and P06748.

Column	Type	Description
ancestry	TEXT	Population ancestry code. Partition column.
protein_id	TEXT	UniProt accession. Partition column.
panel	TEXT	Olink panel containing the assay. Partition column.
chromosome	TEXT	Chromosome on which the variant is located.
position	INT	Variant position in assembly GRCh38 coordinates.
effect_allele	TEXT	Effect allele. Harmonized to equal the dbSNP alt_allele.
other_allele	TEXT	Non-effect allele. Harmonized to equal the dbSNP ref_allele.
beta	FLOAT	Effect size estimate.
standard_error	FLOAT	Standard error of the effect size estimate.
effect_allele_frequency	FLOAT	Frequency of the effect allele in the ancestry group.
neg_log_10_p_value	FLOAT	-log10(p-value). Higher = more significant. Genome-wide significance threshold is ~7.3.
variant_id	TEXT	Unique UKB-PPP variant identifier (e.g. 4:180574458:A:G:imp:v1).
info	FLOAT	Imputation quality score (0-1, higher is better).
n	INT	Sample size.
chi_squared	FLOAT	Chi-squared test statistic.

proteins

Column	Type	Description
partition	INT	Dummy partition column to enable compatibility with Cloudflare R2 SQL API. Ignore.
protein_id	TEXT	UniProt accession (e.g. P43220, P02649), or multiple accessions joined by underscore for multi-chain protein complexes.
protein_name	TEXT	Full protein name (e.g. Glucagon-like peptide 1 receptor). NOT a gene symbol — do not search for gene names here.
uniprot_accessions	TEXT[]	List of constituent UniProt accessions. A single-element list, or a multi-element list for multi-chain protein complexes. Cannot be filtered in queries (array type).

variants

Note

All load_dataset() queries must filter on partition column: chromosome.

Column	Type	Description
variant_id	TEXT	Unique UKB-PPP variant identifier (e.g. 4:180574458:A:G:imp:v1).
rsid	TEXT, nullable	dbSNP rsID (e.g. rs123456). May be null for novel variants.
chromosome	TEXT	Chromosome on which the variant is located. Partition column.
position_grch37	INT	Variant position in assembly GRCh37 coordinates.
position_grch38	INT	Variant position in assembly GRCh38 coordinates.
effect_allele	TEXT	Effect allele of pQTL results. Harmonized to equal the dbSNP alt_allele.
other_allele	TEXT	Non-effect allele of pQTL results. Harmonized to equal the dbSNP ref_allele.
strand	TEXT, nullable	Strand of the variant.
is_palindromic	BOOLEAN	Variant alleles are A/T or C/G. These require specific treatment during harmonization.
dbsnp_build	TEXT	Version of dbSNP used to harmonize variant alleles (e.g. b156).
in_dbsnp	BOOLEAN	Variant has a match in dbSNP (allowing for swapping of alleles).
was_flipped	BOOLEAN, nullable	Variant was flipped relative to source UKB-PPP pQTL files during harmonization.