Abstract

Main

Understanding the role that genetics has in phenotypic and disease variation, and its potential interactions with other factors, is crucial for a better understanding of human biology. It is hoped that this will lead to more successful drug development¹, and potentially to more efficient and personalized treatments. As such, a key component of the UK Biobank resource has been the collection of genome-wide genetic data on every participant using a purpose-designed genotyping array². An interim release of genotype data on approximately 150,000 UK Biobank participants in May 2015³ has already facilitated numerous studies^4–6.

In this paper, we summarize the existing and planned content of the phenotype resource and describe the genetic dataset on the full 500,000 participants. To facilitate its wider use, we applied a range of quality control procedures and conducted a set of analyses that reveal properties of the genetic data—such as population structure and relatedness—that can be important for downstream analyses. In addition, we estimated haplotypes and imputed genotypes into the dataset that increases the number of testable variants by more than 100-fold to approximately 96 million variants. We also imputed classical allelic variation at 11 human leukocyte antigen (HLA) genes, and replicated signals of known associations between HLA alleles and many common diseases. We describe tools that allow efficient genome-wide association studies (GWAS) of multiple traits and fast phenome-wide association studies, which work together with a new compressed file format that has been used to distribute the dataset. As a further check of the genotyped and imputed datasets, we performed a test-case genome-wide association scan on a well-studied human trait, standing height.

The UK Biobank

A wide variety of phenotypic information as well as biological samples have been collected for each of the approximately 500,000 UK Biobank participants (Fig. 1). At recruitment, participants provided electronic signed consent, answered questions on socio-demographic, lifestyle and health-related factors, and completed a range of physical measures (see Extended Data Table 1). They also provided blood, urine and saliva samples, which were stored in such a way as to allow many different types of assay to be performed (for example, genetic, proteomic and metabonomic analyses)⁷. Once recruitment was fully underway, further enhancements were introduced to the assessment visit, including a range of eye measures, an electrocardiograph test, arterial stiffness and a hearing test.

Open in a separate window

Fig. 1

Summary of the UK Biobank resource and genotyping array content.

Summary of the major components of the UK Biobank resource. See Extended Data Table 1 for more details. The figure also shows a schematic representation of the different categories of content on the UK Biobank Axiom genotype array. Numbers indicate the approximate count of markers within each category, ignoring any overlap. A more detailed description of the array content is available in the UK Biobank Axiom Array Content Summary².

Extended Data Table 1

Types and dates of data collection in UK Biobank

Open in a separate window

Types and dates of data collection in UK Biobank

^aThe baseline visit (including the touchscreen questionnaire, physical measures and biological sampling) was repeated approximately 5 years later (2012–2013) in a subset of 20,000 participants and in those who attended an imaging assessment centre (2014–2022).

^bIncludes 70,000 participants who completed the diet online questionnaire at the end of the recruitment visit⁴⁸.

^cA repeat assessment of physical activity on four occasions over a 12-month period is being collected on 2,5000 of these participants (2018–2019).

^dBiochemistry markers were measured in the baseline sample for 500,000 participants and in the repeat assessment sample for 20,000 participants. The urinary biomarkers were made available in 2016; the serum and red blood cell markers available are pending (at the time of press).

^eThe imaging study includes brain, heart and body MRI, carotid ultrasound and 12-lead ECG scan and a full-body dual energy X-ray absorptiometry scan, plus a repeat of the baseline assessment (including biological sampling). Repeat imaging in a subset of participants is expected to start in 2019.

^fData are currently available for 25,000 participants, with the remaining 75,000 participants to attend over the next few years. See Supplementary Table 1 for further information about these data types.

The baseline information has been, and will continue to be, extended in several ways. For example, repeat assessments are planned to be conducted in subsets of the cohort every few years, to enable calibration of measurements, adjustment for regression dilution, and estimation of longitudinal change. Objective measures of physical activity have also been collected (using a tri-axial accelerometer) in 100,000 participants in 2013–2014⁸ with repeated measures being collected over a period of a year (on a seasonal basis) from 2,500 of these participants. A multi-modal imaging assessment is currently underway, which comprises magnetic resonance imaging (MRI) of the brain⁹, heart¹⁰ and body, carotid ultrasound¹¹ and a whole body dual-energy X-ray absorptiometry of the bones and joints¹². Data collection started in 2014 and is anticipated to take 7–8 years to achieve imaging for 100,000 participants in dedicated imaging assessment centres across the United Kingdom, with repeat imaging measures being planned for a subset of participants.

All participants provided consent for follow-up through linkage to their health-related records. As of May 2018, there were over 14,000 deaths, 79,000 participants with cancer diagnoses, and 400,000 participants with at least one hospital admission. Considerable efforts are now underway to incorporate data from a range of other national datasets including primary care, screening programmes, and disease-specific registries, as well as asking participants directly about health-related outcomes through online questionnaires (see Extended Data Table 1). Efforts are also underway to develop scalable approaches that can characterize in detail different health outcomes by cross-referencing multiple sources of coded clinical information¹³.

Measurements for a wide range of biochemical markers of key interest to the research community have also been carried out, including those that have known associations with disease (for example, lipids for vascular disease and sex hormones for cancer), diagnostic value (for example, HbA_1c for diabetes and rheumatoid factor for arthritis), or the ability to characterize phenotypes not otherwise well assessed (for example, biomarkers for renal and liver function).

UK Biobank is an open-access resource that encourages researchers from around the world, including those from the academic, charity, public and commercial sectors, to access the data for any health-related research that is in the public interest.

Whole-genome genotyping

The UK Biobank genetic data contains genotypes for 488,377 participants. These were assayed using two very similar genotyping arrays. A subset of 49,950 participants involved in the UK Biobank Lung Exome Variant Evaluation (UK BiLEVE) study were genotyped at 807,411 markers using the Applied Biosystems UK BiLEVE Axiom Array by Affymetrix (now part of Thermo Fisher Scientific), which is described elsewhere⁶. Following this, 438,427 participants were genotyped using the closely related Applied Biosystems UK Biobank Axiom Array (825,927 markers) that shares 95% of marker content with the UK BiLEVE Axiom Array. The marker content of the UK Biobank Axiom array was chosen to capture genome-wide genetic variation (single nucleotide polymorphism (SNPs) and short insertions and deletions (indels)), and is summarized in Fig. 1. Many markers were included because of known associations with, or possible roles in, disease. The array also includes coding variants across a range of minor allele frequencies (MAFs), including rare markers (<1% MAF); and markers that provide good genome-wide coverage for imputation in European populations in the common (>5%) and low frequency (1–5%) MAF ranges. Further details of the array design are in the UK Biobank Axiom Array Content Summary².

DNA was extracted from stored blood samples that had been collected from participants on their visit to a UK Biobank assessment centre. Genotyping was carried out by Affymetrix Research Services Laboratory in 106 sequential batches of approximately 4,700 samples (see Methods, Supplementary Table 12). Affymetrix applied a custom genotype calling pipeline and quality filtering optimized for biobank-scale genotyping experiments and the novel genotyping arrays, which contain markers that had not been previously typed using Affymetrix technology (see Methods). This resulted in a set of genotype calls for 489,212 samples at 812,428 unique markers (biallelic SNPs and indels) from both arrays, with which we conducted further quality control and analysis (Extended Data Table 2).

Extended Data Table 2

The number of markers and samples by genotyping array at main stages of the UK Biobank genotyping experiment

Open in a separate window

The number of markers and samples by genotyping array at main stages of the UK Biobank genotyping experiment

‘Data delivery from Affymetrix’ refers to the data produced by Affymetrix after applying their filtering (Supplementary Information). ‘Released data’ refers to the released genotype data, after applying quality control measures, as detailed in sections S2 and S3 of the Supplementary Information.

Our quality control pipeline was designed specifically to accommodate the large-scale dataset of ethnically diverse participants, genotyped in many batches, using two slightly different arrays, and which will be used by many researchers to tackle a wide variety of research questions. Participants reported their ethnic background by selecting from a fixed set of categories¹⁴. Although most (94%) individuals report their ethnic background as within the broad-level group ‘white’, there are still approximately 22,000 individuals with a self-reported ethnic background originating outside Europe (Extended Data Table 3). We used approaches based on principal component analysis (PCA) to account for population structure in both marker and sample-based quality control (see Methods).

Extended Data Table 3

Counts and proportions of self-reported ethnic groups among 488,377 genotyped UK Biobank participants

Open in a separate window

Counts and proportions of self-reported ethnic groups among 488,377 genotyped UK Biobank participants

Categories of self-reported ethnic background (UK Biobank data field 21000) and broader-level ethnic groups are shown here to reflect the two-layer branching structure of the ethnic background section in the UK Biobank touchscreen questionnaire¹⁴. Participants first picked one of the broader-level ethnic groups (for example, ‘white’), and were then prompted to select one of the categories within that group (for example, ‘Irish’). The broader-level groups are also shown here as an ethnic background category (‘white’ in column two) because a small proportion of participants only responded to the first question. In this table, we also combine the category ‘other ethnic group’ with an aggregated non-response category ‘not stated’, which includes all participants who did not know their ethnic group, or stated that they preferred not to answer, or did not answer the first question.

To identify poor quality markers, we used statistical tests designed primarily to check for consistency across experimental factors, such as array or batch (see Methods; Extended Data Table 4). As a result of these tests, we set to missing 0.97% of all the genotype calls made by Affymetrix. We identified poor quality samples using the metrics of missing rate and heterozygosity adjusted for population structure (Extended Data Fig. 1), as extreme values in one or both of these metrics can be indicators of poor sample quality due to, for example, DNA contamination¹⁵. We identified 968 such samples (0.2%), and provide this list to researchers.

Open in a separate window

Extended Data Fig. 1

Summary of sample-based quality control.

a–c, The three plots show heterozygosity and missing rates, which we used to flag poor quality samples (n=488,377 samples). Panels a and b show heterozygosity for each sample before and after, respectively, correcting for ancestral background using principal components. The symbols (shapes and colours) indicate the self-reported ethnic background of each participant. Panel c shows the set of 968 samples we flagged as outliers (in red), and all other samples (in black), with shapes the same as for the other two plots. The vertical line shows the threshold we used to call samples as outliers on missing rate. In all plots missing rate data are transformed to the logit scale, but with the axis annotated with the original values.

Extended Data Table 4

Failure rates for six marker-based quality tests

Open in a separate window

Failure rates for six marker-based quality tests

For all numbered tests, a marker (or marker within a batch) was set to missing if the test yielded P<10⁻¹², except in the case of test 6, for which a marker was set to missing if the test yielded <95% concordance. See Supplementary Information for details of each test (n=463,844 samples). The total is not equal to the sum of all tests because it is possible for a marker to fail more than one test. Because the two arrays contain slightly different sets of markers, the total number of genotype calls used to compute the fractions is:

N_ukbb L_ukbb + N_ukbl L_ukbl, in which N and L refer to the numbers of markers and samples typed on the UK Biobank Axiom array (ukbb) and samples typed on the UK BiLEVE Axiom array (ukbl) within the Affymetrix data delivery (see Supplementary Table 1).

^aThe array effect test was applied across all batches and only for markers present on both arrays, so we simply report the total number of markers that failed this test.

^bThe discordance test was applied across all batches, but not all markers are present on both arrays. The first value is the number of unique markers on the UK BiLEVE Axiom array that failed this test, and the second is for markers on the UK Biobank Axiom array.

Mismatches between self-reported sex of each individual, and sex inferred from the relative intensity of markers on the Y and X chromosomes¹⁶, can be used as a way to detect possible sample mishandling or other types of clerical error. In a dataset of this size, some such mismatches would be expected due to transgender or intersex individuals, or instances of rare genetic variation, such as sex-chromosome aneuploidies¹⁷. Using information in the measured intensities of chromosomes X and Y (see Methods), we identified a set of 652 (0.134%) individuals with sex chromosome karyotypes that were putatively different from XY or XX (Fig. 2d, Supplementary Table 2).

Open in a separate window

Fig. 2

Summary of genotype data quality and content.

All plots show properties of the UK Biobank genotype data after applying quality control. a, MAF distribution based on all samples (805,426 markers). The inset shows rare markers only (MAF<0.01). b, The distribution of the number of batch-level quality control (QC) tests that a marker fails (see Methods). For each of four MAF ranges, we show the fraction of markers that fail the specified number of batches. c, Comparison of MAF in UK Biobank with the frequency of the same allele in ExAC, among the European-ancestry participants within each study (Supplementary Information). This analysis used 91,298 overlapping markers. Each hexagonal bin is coloured according to the number of markers falling in that bin (log₁₀ scale). The dashed red line shows x=y. The markers with very different allele frequencies seen on the top, bottom and left-hand sides of the plot comprise approximately 300 markers. This is 0.3% of all markers in the comparison (see Supplementary Information for discussion). d, Mean log₂ ratios (L₂R) on X and Y chromosomes for each sample, indicating probable sex chromosome aneuploidy (see Methods). There are 652 samples with a probable sex chromosome aneuploidy (indicated by crosses). Locations of clusters of individuals with different putative karyotypes are indicated by Greek symbols: λ=X0 (or mosaic XX/X0), θ=XXX, α=XXY, and π=XYY. Counts of individuals in these regions are given in Supplementary Table 2. The colours indicate different combinations of self-reported sex, and sex inferred by Affymetrix (from the genetic data). For almost all samples (99.9%), the self-reported and the inferred sex are the same, but for a small number of samples (378) they do not match (see Supplementary Information for discussion).

The application of our quality control pipeline resulted in the released dataset of 488,377 samples and 805,426 markers from both arrays with the properties shown in Fig. 2a–c. A set of 588 pairs of experimental duplicates show very high genotype concordance, with mean 99.87% and minimum 99.39% of genotypes identical (Supplementary Fig. 13). We compared allele frequencies among UK Biobank participants with European ancestry to those estimated from an independent source, the Exome Aggregation Consortium (ExAC) database¹⁸ at a set of 91,298 overlapping markers. We do not expect allele frequencies in the two studies to match exactly owing to subtle differences in the ancestral backgrounds of the individuals in each study, as well as differences in the sensitivity and specificity of the two technologies (exome sequencing and genotyping arrays). A small number of markers (around 300) have very different allele frequencies (see Supplementary Information section 2.4). This could be due to non-working probesets on the UK Biobank arrays or possibly annotation error on the UK Biobank arrays or in ExAC, or mapping errors in the sequence data in regions of more complex variation. Despite this, overall the allele frequencies are encouragingly similar (r²=0.93) (Fig. 2c; Supplementary Fig. 4).

More than 110,000 rare markers (MAF<0.01 in UK Biobank) were included on the two arrays used for the UK Biobank cohort². Variants occurring at very low frequencies present a particular challenge for genotype calling using array technology. It can be challenging to distinguish a sample that genuinely has the minor allele, from one in which the intensities are in the tails of the distribution of those in the major homozygote cluster (Extended Data Fig. 2). A larger fraction of rare markers fail quality control tests compared to low frequency and common markers, but 84% still pass in all batches (Fig. 2b). We recommend researchers visually inspect cluster plots, similar to Supplementary Fig. 2, for markers of interest using a utility such as Evoker (https://github.com/wtsi-medical-genomics/evoker), especially for rare markers.

Open in a separate window

Extended Data Fig. 2

Examples of intensity data and genotype calls for markers of different allele frequencies.

Each sub-figure shows intensity data for a single marker within six different batches. Batches labelled with the prefix ‘UKBiLEVEAX’ contain only samples typed using the UK BiLEVE Axiom array, and those with the prefix ‘batch’ contain only samples typed using the UK Biobank Axiom array. Each point represents one sample and is coloured according to the inferred genotype at the marker. The x and y axes are transformations of the intensities for probe sets targeting each of the alleles ‘A’ and ‘B’ (see Supplementary Information for definition of probe set). The ellipses indicate the location and shape of the posterior probability distribution (two-dimensional multivariate normal) of the transformed intensities for the three genotypes in the stated batch. That is, each ellipse is drawn such that it contains 85% of the probability density. See Affymetrix Axiom Genotyping Solution Data Analysis Guide¹⁶ for more details of Affymetrix genotype calling. The MAF of each of the markers is computed using all samples in the released UK Biobank genotype data. a, A marker with a MAF of 0.077 with well-separated genotype clusters. b, Intensities for a marker with a MAF of 0.00092 with well-separated genotype clusters. As would be expected under Hardy–Weinberg equilibrium, there are no instances of samples with the minor homozygote genotype. c, Intensities for a marker with a MAF of 0.00066, and in which the heterozygote cluster is not well separated from the large major homozygote cluster in some batches, making it more difficult to call the heterozygous genotypes confidently.

Ancestral diversity and cryptic relatedness

The genotype data provide a unique opportunity to study the diverse ancestral origins (Extended Data Table 3) of UK Biobank participants. Accounting for the ancestral background is essential both for epidemiological studies and genetic analyses, such as GWAS¹⁹. We used PCA to measure population structure within the UK Biobank cohort (see Methods). Figure 3a shows results for the first four principal components plotted in consecutive pairs (see also Extended Data Fig. 3 and Supplementary Figs. 6, 7). As expected, individuals with similar principal component scores have similar self-reported ethnic backgrounds. For example, the first two principal components separate out individuals with sub-Saharan African ancestry, European ancestry and east Asian ancestry. Individuals who self-report as mixed ethnicity tend to fall on a continuum between their constituent groups. Further principal components capture population structure at sub-continental geographic scales (Extended Data Fig. 3). Our PCA revealed population structure within the most common ethnic background category (88.26%), ‘British’ within the broader-level group ‘white’ (Supplementary Fig. 8). We used a combination of self-reported ethnic background and PCA results to provide researchers with a list of 409,728 individuals (84%) who have very similar ancestral backgrounds relative to the full cohort (see Methods).

Open in a separate window

Fig. 3

Ancestral diversity and familial relatedness.

a, Each point represents a UK Biobank participant (n=488,377 samples) and is placed according to their principal component (PC) scores in each of the top four principal components. Colours and shapes indicate the self-reported ethnic background of each individual. See Extended Data Table 3 for proportions in each category. b, Distribution of the number of relatives that participants have in the UK Biobank cohort. The height of each bar shows the count of participants (log₁₀ scale) with the stated number of relatives. The colours indicate the proportions of each relatedness class within a bar. c, Examples of family groups within the UK Biobank cohort. Points represent participants, and coloured lines between points indicate their inferred relationship (for example, blue lines join full siblings). The integers show the total number of family networks in the cohort (if more than one) with that same configuration, ignoring third-degree pairs.

Open in a separate window

Extended Data Fig. 3

Mean principal component scores for each self-reported country of birth.

Each column shows one principal component and each element is the mean principal component score for individuals born in the labelled country, scaled by the standard deviation of the scores for that principal component. Elements in each column are only coloured if the country has a non-zero coefficient (P<10⁻⁵; two-sided t-test) in a linear model with country of birth as predictor and principal component scores as outcome (n=487,848 samples). Countries (rows) have been ordered using hierarchical clustering (‘hclust’ function in R). The symbols next to each country label indicate the most common ethnic background category among the participants born in that country. For example, the most common self-reported ethnic background of participants born in Sri Lanka is ‘Any other Asian background’. Countries with fewer than 20 individuals born there were excluded from this analysis.

Close relationships (for example, siblings) among UK Biobank participants were not recorded during the collection of other phenotypic information. This information can be important for epidemiological analyses²⁰, as well as in GWAS²¹. We used the genetic data to identify related individuals by estimating kinship coefficients for all pairs of samples, and report coefficients for pairs of relatives who we infer to be third-degree relatives or closer (see Methods). A total of 147,731 UK Biobank participants (30.3%) are inferred to be related (third degree or closer) to at least one other person in the cohort, and form a total of 107,162 related pairs (Extended Data Table 5). This is a surprisingly large number, and it is not driven solely by an excess of third-degree relatives. For example, the number of sibling pairs (22,666) is roughly twice as many as would theoretically be expected in a random sample (of this size) of the eligible UK population, after taking into account typical family sizes (Supplementary Table 4). The larger than expected number of related pairs could be explained by sampling bias due to, for example, an individual being more likely to agree to participate because a family member was also involved. Furthermore, if, as seems plausible, related individuals cluster geographically rather than being randomly located across the UK, the recruitment strategies of the UK Biobank assessment centres²² will naturally tend to oversample related individuals.

Extended Data Table 5

Summary of related pairs (third-degree relatives or closer) for the full UK Biobank cohort

Open in a separate window

Summary of related pairs (third-degree relatives or closer) for the full UK Biobank cohort

Counts are derived from the kinship coefficients (see Methods). The count of monozygotic twins is after excluding samples identified as duplicates (Supplementary Information).

Pairs of related individuals within the UK Biobank cohort form networks of related individuals. In most cases, these are of size two, but there are also many groups of size three or larger in the cohort (Fig. 3b), even when restricting to second-degree relatives or closer relative pairs. By considering the relationship types and the age and sex of the individuals within each family group, we identified 1,066 sets of trios (two parents and an offspring), which comprise 1,029 unique sets of parents and 37 quartets (two parents and two children).

There are 172 family groups with 5 or more individuals that are second-degree relatives or closer (Fig. 3c). One such group has 11 individuals who are all second-degree relatives of each other (half-siblings, grandparent/grandchild, or avuncular). Because all of the 55 pairs are second-degree relatives, at least 10 of them must be half-siblings with the same shared parent (see Supplementary Material). We confirmed that the shared parent must be their father because they do not all carry the same mitochondrial alleles, and the males all have the same Y chromosome alleles (data not shown).

Haplotype estimation and genotype imputation

We estimated haplotypes for the full cohort (pre-phasing), followed by haploid imputation²³. For the pre-phasing step, we only used markers present on both the UK BiLEVE and UK Biobank Axiom arrays. We removed markers that failed quality control in more than one batch, had a greater than 5% overall missing rate, and had a MAF of less than 0.0001. We removed samples that were identified as outliers for heterozygosity and missing rate. These filters resulted in a dataset with 670,739 autosomal markers in 487,442 samples. Phasing on the autosomes was carried out using SHAPEIT3²⁴ (see Methods and https://jmarchini.org/software/). The 1000 Genomes phase 3 dataset²⁵ was used as a reference panel, predominantly to help with the phasing of samples with non-European ancestry. In a separate experiment that leveraged phase inferred from mother–father–child trios, we estimated a median phasing switch error rate of 0.229% (see Methods).

We used the Haplotype Reference Consortium (HRC)²⁶ data as the main imputation reference panel because it consisted of the largest available set (64,976) of broadly European haplotypes at 39,235,157 SNPs. Supplementary Fig. 15 shows the results of a separate imputation experiment that shows that the HRC panel produces better imputation performance than the UK10K panel, especially at lower allele frequencies, and that the UK Biobank Axiom array performs favourably compared to other commercially available arrays.

We also imputed the UK Biobank using the merged UK10K and 1000 Genomes phase 3 reference panels²⁷, which has 87,696,888 bi-allelic markers. We combined this imputed data with that from the HRC panel, using the HRC imputation when a SNP was present in both panels. Imputation was carried out with the IMPUTE4 program (https://jmarchini.org/software/), which is a re-coded version of the haploid imputation functionality implemented in IMPUTE2²³ (see Methods). The result of the imputation process is a dataset with 93,095,623 autosomal SNPs, short indels and large structural variants in 487,442 individuals. We imputed an additional 3,963,705 markers on the X chromosome (Methods). The SNP database (dbSNP) reference SNP (rs) IDs were assigned to as many markers as possible using reference SNP ID lists available from the UCSC genome annotation database for the GRCh37 assembly of the human genome (http://hgdownload.cse.ucsc.edu/goldenpath/hg19/database/).

Extended Data Fig. 4 shows the distribution of information scores on all markers in the imputed dataset. An information score of α in a sample of M individuals indicates that the amount of data at the imputed marker is approximately equivalent to a set of perfectly observed genotype data in a sample size of αM. The figure illustrates that most markers above 0.1% frequency have high information scores. Previous GWAS have tended to use a filter on information around 0.3 that roughly corresponds to an effective sample size of approximately 150,000. Thus, it may be possible to reduce the information score threshold and still obtain good power to detect associations.

Open in a separate window

Extended Data Fig. 4

Distribution of information scores at autosomal markers in the imputed dataset.

The top left graph shows the full distribution of the information scores. The remaining panels show distributions in tranches of MAF; MAF > 5%, 1% ≤ MAF<5%, 0.1%≤MAF<1%, 0.01%≤MAF<0.1% and 0.001%≤MAF<0.01%.

We developed a new BGEN file format (v1.2; http://www.well.ox.ac.uk/~gav/bgen_format/bgen_format.html) and software library (BGEN; https://bitbucket.org/gavinband/bgen) to provide improved data compression, the ability to store phased haplotype data and random access to the data via use of a separate index file. Using this new format, the full imputed files require 2.1 Tb of file space. A new program (BGENIE; https://jmarchini.org/software) was built using the BGEN library to carry out fast multi-trait GWAS and phenome-wide association studies²⁸ (see Supplementary Information).

Imputation of classical HLA alleles

The major histocompatibility complex (MHC) on chromosome six is the most polymorphic region of the human genome and contains the largest number of genetic associations to common diseases²⁹. We imputed HLA types at two-field (also known as four-digit) resolution for 11 classical HLA genes (HLA-A, HLA-B, HLA-C, HLA-DRB1, HLA-DRB3, HLA-DRB4, HLA-DRB5, HLA-DQA1, HLA-DQB1, HLA-DPA1 and HLA-DPB1) using the HLA*IMP:02 algorithm with a multi-population reference panel (Supplementary Tables 5 and 6)³⁰ and validated the accuracy using a cross-validation experiment. In a typical use, case accuracy was estimated at better than 96% across all loci (see Methods and Supplementary Tables 7, 8).

To demonstrate the utility of the HLA imputation, we performed association tests for diseases known to have HLA associations. We analysed 409,724 individuals in the white British ancestry subset (see Methods) and focused on 11 self-reported immune-mediated diseases with known HLA associations. For each disease in our analysis, we identified the HLA allele with the strongest evidence of association. In all cases these were consistent with previous reports (see Methods and Supplementary Table 9). We further replicated independent HLA associations in a single disease study of multiple sclerosis (MS) susceptibility by the International Multiple Sclerosis Genetics Consortium (IMSGC)³¹. Here we observed evidence of association and effect size estimates for HLA alleles that are concordant in direction and relative magnitude with those found in the IMSGC study, although in 11 out of 14 cases this was closer to 1, consistent with regression dilution bias arising from a low rate of phenotypic error (Table 1).

Table 1

Association between HLA alleles and MS in UK Biobank and IMSGC cohort

HLA allele	Test	UK Biobank		IMSGC
		OR (95% CI)	P value	OR (95% CI)	P value
HLA-DRB1*15:01	Additive effect	3.16 (2.81–3.54)	2.58×10−85	3.92 (3.74–4.12)	<1×10−600
	Homozygote correction	0.67 (0.52–0.87)	2.32×10−3	0.54 (0.47–0.61)	8.50×10−22
HLA-A*02:01	Additive effect	0.69 (0.62–0.78)	2.30×10−10	0.67 (0.64–0.70)	7.80×10−70
	Homozygote correction	1.20 (0.89–1.62)	2.41×10−1	1.26 (1.13–1.41)	3.30×10−05
HLA-DRB1*03:01	Additive effect	1.21 (1.06–1.37)	3.39×10−3	1.16 (1.10–1.22)	3.50×10−08
	Homozygote correction	2.12 (1.53–2.94)	6.84×10−6	2.58 (2.19–3.03)	1.30×10−30
HLA-DRB1*13:03	Additive effect	2.10 (1.54–2.85)	2.36×10−6	2.62 (2.32–2.96)	6.20×10−55
HLA-DRB1*08:01	Additive effect	1.56 (1.21–2.01)	6.13×10−4	1.55 (1.42–1.69)	1.00×10−23
HLA-B*44:02	Additive effect	0.86 (0.74–0.98)	2.94×10−2	0.78 (0.74–0.83)	4.70×10−17
HLA-B*38:01	Additive effect	0.29 (0.13–0.65)	2.55×10−3	0.48 (0.42–0.56)	8.00×10−23
HLA-B*55:01	Additive effect	0.99 (0.75–1.31)	9.47×10−1	0.63 (0.55–0.73)	6.90×10−11
HLA-DQA1*01:01	Additive effect in the presence of HLA-DRB1*15:01	0.71 (0.56–0.90)	5.33×10−3	0.65 (0.59–0.72)	1.30×10−17
HLA-DQB1*03:02	Dominant effect	1.07 (0.92–1.25)	3.71×10−1	1.30 (1.23–1.37)	1.80×10−22
HLA-DQB1*03:01	Allelic interaction with HLA-DQB1*03:02	0.8 (0.53–1.20)	2.81×10−1	0.60 (0.52–0.69)	7.10×10−12

Open in a separate window

Evidence for association between HLA alleles and MS in UK Biobank compared to the IMSGC cohort. The UK Biobank association tests involved 1,501 self-reported cases and 409,724 controls; the IMSGC cohort involved 17,465 cases and 30,385 controls³¹. Thus, the UK Biobank analysis had significantly lower power than the IMSGC analysis, which is reflected in the reported P values and larger confidence interval (CI) estimates for the odds ratios (OR). Effect sizes for the UK Biobank were estimated jointly using the logistic regression model of the MHC reported by the IMSGC (with the exception of the two SNPs rs9277565 and rs2229029). As in the IMSGC analysis, the homozygote correction test indicates a departure from additivity. That is, if the odds ratio is <1 then the homozygous effect is smaller than under the additivity assumption and bigger if it is >1. Reported P values were calculated using the Wald test.

GWAS for standing height

To assess the potential of the directly genotyped and imputed data, we conducted a GWAS for standing height using 343,321 unrelated, European-ancestry UK Biobank participants (see Methods). We compared our results to a non-overlapping meta-analysis of 253,288 individuals of European ancestry carried out by the Genetic Investigation of Anthropometric Traits (GIANT) Consortium³².

Reassuringly, the pattern of association signals is similar in both the UK Biobank and GIANT results (Fig. 4a–c), and the Z-scores at associated markers are highly correlated (r²=0.965; Fig. 4e). The gain in power in the UK Biobank cohort is clear, with many loci reaching genome-wide significance (P<5×10⁻⁸) in the UK Biobank but not in the GIANT study (Fig. 4d, Supplementary Fig. 16); and Z-scores for associated markers are systematically higher in UK Biobank (regression slope=1.369, Fig. 4e). Regions of association in the UK Biobank show patterns of signal expected given the linkage disequilibrium structure and recombination rates in the region (see Extended Data Fig. 5 for an example).

Open in a separate window

Fig. 4

Association statistics for human height.

Results (P values) of association tests between human height and genotypes using three different sets of data for chromosome 2. In a–c, P values are shown on the −log₁₀ scale, capped at 50 for visual clarity and uncorrected for multiple comparisons. Markers with −log₁₀(P) > 50 are plotted at 50 on the y axis and shown as triangles rather than dots. Horizontal red lines denote P=5×10⁻⁸. a, Results for published meta-analysis by GIANT³² (n=253,288), with NCBI GWAS catalogue markers superimposed in red (plotted at the reported P values). b, Association statistics (from linear mixed model, see Methods) for UK Biobank markers in the genotype data (n=343,321). c, Association statistics (from linear mixed model, see Methods) for UK Biobank markers in the imputed data (n=343,321). Points coloured pink indicate genotyped markers that were used in pre-phasing and imputation. This means that most of the data at each of these markers comes from the genotyping assay. Black points (the vast majority, ~8 million) indicate fully imputed markers. d, Venn diagram of the results of counting the number of 1-Mb windows with at least one locus with P<5×10⁻⁸ in the GIANT, UK Biobank genotyped and UK Biobank imputed datasets (see Methods). Percentages in brackets are the proportion of the union of such windows across all three data sources (1,215). There were only three windows contained in UK Biobank genotyped data and not the imputed data. e, Comparison of Z-scores in UK Biobank (y axis) and GIANT (x axis). Z-scores were calculated as effect size divided by standard error, but only for markers with P<5×10⁻⁸ in GIANT, for a set of 575 associated regions, which we also used for the credible set analysis (see Methods). The marker with the smallest P value (in GIANT) within each region is highlighted with blue circles. The black dotted line shows x=y, and the red solid line shows the linear regression line estimated on these data. The standard error of the regression coefficient is shown in brackets. Pearson’s correlation was used to calculate the r² value.

Open in a separate window

Extended Data Fig. 5

Example region of association in standing height GWAS.

GWAS association statistics (P values) for standing height focusing on a ~3-Mb region of chromosome 2 that did not reach genome-wide significance in the GIANT (2014) meta-analysis, but did in UK Biobank (linear mixed model; see Methods). The P values shown are not adjusted for multiple testing. Markers genotyped in the UK Biobank are shown as diamonds, and imputed markers as circles. The two markers with the smallest P value for each of the genotyped data and imputed data are enlarged and highlighted with black outlines, and other UK Biobank markers are coloured according to their correlation (r²) with one of these two. That is, genotyped markers with the leading genotyped marker (rs17713396), and imputed markers with the leading imputed marker (rs12714401). Markers with r² values of less than 0.1 are shown as black or green.

To assess the effectiveness of UK Biobank genomic data for fine-mapping within associated loci, we computed 95% credible sets³³ for 575 regions that contain at least one genome-wide significant marker (P<5×10⁻⁸) in both GIANT and the UK Biobank imputed data (see Methods). The number of markers we analysed in the UK Biobank (768,502) is considerably more than in GIANT (106,263), and this affects the resolution of any given associated region (Extended Data Fig. 6a). When considering all markers, the size of the credible set in UK Biobank is usually larger (median size=8) than in GIANT (median size=6), but the proportion of SNPs in the credible set of each region (Extended Data Fig. 6b) is generally smaller in UK Biobank (median proportion=0.010) than in GIANT (median proportion=0.047). By restricting to the markers in both studies (105,421) we find that the size of the 95% credible set is generally smaller in UK Biobank (median size=4) than GIANT (median size=6). The number of 95% credible sets that contain just 1 marker is 123 in UK Biobank and 76 in GIANT.

Open in a separate window

Extended Data Fig. 6

Comparison of fine-mapping in GIANT (2014) and UK Biobank imputed data.

Here we summarize results of our credible set analysis in GIANT (2014) and UK Biobank for 575 genomics regions associated with standing height in both studies (see Methods). A red solid line on a plot indicates where x=y. a, Both plots compare the number of markers in the 95% credible sets in which the size is less than 18 markers in both studies (363 regions in the left-hand plot; 445 in the right-hand plot). b, c, Both plots are from the analysis considering all markers in each study. In b we show, for each region, the proportion of markers used in the analysis for a given study that are in the 95% credible set for that study. The plot contains the same 363 regions as shown in the left-hand plot in a. In c we summarize, for all 575 regions, how much weight our UK Biobank analysis placed on markers that our analysis of GIANT (2014) indicated were important.

Conclusion

The interim release of the genetic data on approximately 150,000 participants in UK Biobank has already facilitated many papers exploring the links between human genetic variation and disease, and their connection with a wide range of environmental and lifestyle factors. The UK Biobank continues to grow with the addition of further phenotypic information and as researchers return the results of their analyses for UK Biobank to share. Online resources are being developed for sharing the results of analyses using UK Biobank data, including the release of GWAS results for thousands of phenotypes (http://www.nealelab.is/uk-biobank) and the Oxford Brain Imaging Genetics server²⁸ (http://big.stats.ox.ac.uk/). We anticipate that the availability of the full genetic data for UK Biobank will result in a further step change in this productive research cycle. The UK Biobank is a powerful example of the immense value that can be achieved from large population scale studies that combine genetics with extensive and deep phenotyping and linkage to health records coupled with a strong data sharing policy. It is likely to herald a new era in which these and related resources drive and enhance understanding of human biology and disease.

Methods

Online content

Any methods, additional references, Nature Research reporting summaries, source data, statements of data availability and associated accession codes are available at 10.1038/s41586-018-0579-z.

Acknowledgements

Author contributions

A.Y., N.A. and M.E. coordinated data availability. Data analysis was performed as follows: quality control analysis (C.B., C.F., D.P. and S.W.), HLA imputation and association testing (A.C., A.M. and D.V), phasing, imputation, file formats and multiple trait analysis (O.D., J.O., G.B., K.S., L.T.E. and J.M.) and GWAS testing (C.B., C.F. and J.M.). Supervision of these activities was provided by G.M., P.D. and J.M. C.B., C.F., A.C., S.L., N.A., G.M., P.D. and J.M. wrote the paper.

Data availability

The genetic and phenotype datasets generated by UK Biobank analysed during the current study are available via the UK Biobank data access process (see http://www.ukbiobank.ac.uk/register-apply/). Detailed information about the genetic data available from UK Biobank is available at http://www.ukbiobank.ac.uk/scientists-3/genetic-data/ and http://biobank.ctsu.ox.ac.uk/crystal/label.cgi?id=100314. The exact number of samples with genetic data currently available in UK Biobank may differ slightly from those described in this paper.

Competing interests

J.M. is a founder and director of Gensci Ltd. P.D., G.M. and S.L. are partners in Peptide Groove LLP. G.M. and P.D. are founders and directors of Genomics Plc. The remaining authors declare no competing financial interests.

Footnotes

References