Heritability of >2,000 traits and disorders in the UK Biobank
As you may know, the UK Biobank recently released data on ~500,000 individuals to the scientific community to advance research on genetics. The Neale Lab has been working to make some basic analyses of that data freely available; you can read more about the overall effort here.
Building on the initial genome-wide association analyses, we’ve now estimated the heritability of each of those traits and disorders, over 2,000 as of this writing. As with the primary analyses, we’re sharing the heritability results for download here, and we have made them available for browsing too.
You can learn more about the heritability analysis in this series of blog posts (links):
Insights from estimates of SNP-heritability for >2,000 traits and disorders in UK Biobank [coming soon].
More technical descriptions are also available:
We view these results as preliminary and subject to change as we continue to refine the analyses as well as the genome-wide association results they are based on. But we hope you’ll agree that they provide an interesting and useful first look at the genetics of the many diverse traits and disorders studied in the UK Biobank.
If you want more information, please contact us at: email@example.com.
Where can I find the heritability results?
What method have you used to estimated heritability?
To enable this quick initial analysis on such large-scale data we’re relying on LD score regression. Full code and technical details are available in the UKBB_ldsc github repository and described in Heritability 501.
Are the primary GWAS results used for this analysis available?
How should I interpret these heritability results?
Very carefully. For starters, we’ve written a couple of primers that cover the concept of heritability and common misconceptions (Heritability 101) and the specific type of heritability that’s being estimated in this analysis (Heritability 201). A longer discussion of the many limitations of this analysis is included in our more technical description of the project (Heritability 501).
We’ll revisit this question in future posts as we explore the results in more detail, but here’s the executive summary: This is a very rough initial analysis and should be treated accordingly. All estimates are for (a) the heritability from common genetic variants (b) of automatically cleaned and transformed phenotypes (c) in a non-random population sample of the UK (d) estimated with a fairly simple statistical model. We expect the results will mostly be useful for generating hypotheses for follow-up (e.g. identifying trends in the estimates for categories of phenotypes, choosing phenotypes with robust statistical evidence for heritability), rather than being a definitive statement about the “true” heritability of a individual phenotype.
Heritability analysis team:
Authored by Raymond Walters with contributions from Claire Churchhouse and Rosy Hosking