Improved polygenic prediction by Bayesian multiple regression on summary statistics

Accurate prediction of an individual’s phenotype from their DNA sequence is one of the great promises of genomics and precision medicine. We extend a powerful individual-level data Bayesian multiple regression model (BayesR) to one that utilises summary statistics from genome-wide association studies (GWAS), SBayesR. In simulation and cross-validation using 12 real traits and 1.1 million variants on 350,000 individuals from the UK Biobank, SBayesR improves prediction accuracy relative to commonly used state-of-the-art summary statistics methods at a fraction of the computational resources. Furthermore, using summary statistics for variants from the largest GWAS meta-analysis (n ≈ 700, 000) on height and BMI, we show that on average across traits and two independent data sets that SBayesR improves prediction R2 by 5.2% relative to LDpred and by 26.5% relative to clumping and p value thresholding.


Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated Our web collection on statistics for biologists contains articles on many of the points above.

Software and code
Policy information about availability of computer code Data collection UK Biobank data were downloaded from repository under application number 12514. Quality control was performed using the PLINK v1.90b3.41 software. Ancestry assignment and relatedness assignment was performed using principal component projection in GCTA (v1.90.0beta). Principal components used for covariate adjustment for summary statistics generation were calculated using flashPCA version 2. Other data set QC had been previously performed in other work with the QC parameters described in the Materials and Methods data section. PLINK v1.90b3.41 and GCTA (v1.90.0beta) were the primary software used to process these data.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

October 2018
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability This study makes use of data from dbGaP ARIC, HRS and (accessions: phs000090, phs000091 and phs000674.v2.p2)), UK10K project (EGA accessions: EGAS00001000108 and EGAS00001000090), and UK Biobank Resource (application number: 12514). Estonian Biobank data were accessed via request from the Estonian Genome Center through data release procedures described at https://www.geenivaramu.ee/en/biobank.ee/data-access.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
For the simulation studies a random subset of 100,000 individuals from the 348,580 unrelated European ancestry individuals from the UK Biobank were chosen. The simulation studies were designed to be a within data comparison between the new proposed method and existing methods. From experience this sample size is sufficient (more than) to produce accurate predictors. For the cross-validation analyses the full set of 348,580 unrelated European ancestry individuals from the UKB were used for summary statistics generation. This sample size was chosen to observe the maximum prediction accuracy that could be achieved using unrelated individuals in the UKB data set. In the across Biobank prediction analyses the full set of 456,426 UKB European ancestry individuals from the UKB were used. For out of sample prediction, the HRS data set consisted of 8,552 unrelated individuals, which was the maximum number available. For the Estonian Biobank 32,594 individuals genotyped on the Global Screening Array were used, which was the maximum available at the time of analysis.  Figure S6). This left 932,969 and 909,293 variants with summary information for height and BMI respectively. These sets of variants were also used in the LDpred and RSS analyses.

Replication
The major experimental findings include the validation of the the newly proposed method to be more accurate at polygenic prediction at a much smaller computational cost. The breadth of scenarios and real data analyses are sufficient, we believe, evidence for reviewers to assess these conclusions.
Randomization For each of the genome-wide association studies age, sex and 10 principal components were adjusted for. These covariates are standard in these types of genetic analyses.

Blinding
Blinding in population data collection is not concern as no treatment is being investigated.
Reporting for specific materials, systems and methods All data we access from previous studies that detail the population characteristics. The data and the agreement numbers have been acknowledged in the manuscript.

Recruitment
All data we access from previous studies that detail the data recruitment. The data and the agreement numbers have been acknowledged in the manuscript.

Ethics oversight
All data we access from previous studies that detail the ethics oversight, which is mentioned in detail in the data methods section. The data and the agreement numbers have been acknowledged in the manuscript.
Note that full information on the approval of the study protocol must also be provided in the manuscript.