SNP Heritability is estimated are from LD Score Regression (LDSR; Bulik-Sullivan et al. 2015 or on pubmed) using ldsc and MTAG
Full code is available on the Github repo
The Neale Lab blog has a full technical discussion on the limitations of this analysis and recommendations for intepreting the results, as well as more general details about our analysis of UK Biobank.
Results on this site are very loosely filtered to report only phenotypes with effective N > 200. Where appropriate, stricter filters are applied as indicated. Results for phenotypes with effective N < 200 can still be found in the dropbox downloads folder.
We follow the conventional approach of restricting to GWAS results from HapMap3 sites passing MAF > 0.01 and INFO > 0.9 as input, but we’ve reimplemented this to define a single passing list in UKB rather than running munge_sumstats.py from ldsc with the w_hm3.snplist reference file.
We’ve run both univariate and partitioned heritability analyses. For reasons that will be discussed in a later post we’re currently presenting the partitioned heritability LDSR results as the primary analysis in the summary browser, but the univariate results are also available here.
The univariate LDSR analysis was run with default settings using precomputed LD scores from 1000 Genomes European ancestry samples (i.e. ./eur_w_ld)
The partitioned LDSR analysis was run using the v1.1 of the Baseline-LD annotations described by Gazal et al. 2017 (also on biorxiv) computed from 1000 Genomes Phase 3 data from European ancestry populations and corresponding allele frequencies (available from the ldsc reference downloads page). Default settings were used, with the exception of removing the maximum \(\chi^2\) filter due to the extreme sample size of UK Biobank.
LDSR was run using the implementation in MTAG, which provides a convenient interface to ldsc from within python rather than via the command line.
All analyses were run on the Google Cloud Platform, with the assistance of Hail, cloudtools, and many other tools.
This site has been generated with the assistance of R Markdown with plotly, DT, and crosstalk, and is hosted via Github Pages.