Reconciling S-LDSC and LDAK functional enrichment estimates - PubMed (original) (raw)

Reconciling S-LDSC and LDAK functional enrichment estimates

Steven Gazal et al. Nat Genet. 2019 Aug.

No abstract available

PubMed Disclaimer

Conflict of interest statement

Competing Financial Interests Statement. The authors declare no conflict of interest.

Figures

Figure 1:

Figure 1:. Likelihood comparison of different models of per-SNP heritability.

We report the change in log likelihood compared to the LDAK model (ΔLL) of five other per-SNP heritability models, summed across 16 independent UK Biobank traits (N = 20K). All six models include one heritability parameter that is maximized in-sample when estimating the likelihood; any other parameters are maximized out-of-sample. (a) Analyses using M = 2,835,699 well-imputed 1000G SNPs (as in ref. ). (b) Analyses using M = 4,631,901 well-imputed HRC SNPs. Numbers between parentheses in figure legends indicate the number of traits with ΔLL > 0. The M = 4.6 million well-imputed HRC SNPs consistently attained higher likelihoods than the M = 2.8 million well-imputed 1000G SNPs in comparisons using the same model. Further details and numerical results are provided in the Supplementary Note.

Figure 2:

Figure 2:. Comparison of functional enrichment estimates in analyses of UK Biobank traits.

For 28 functional annotations, we report functional enrichment estimates of S-LDSC with the baseline-LD model (N = 434K) (a), the LDAK method (N = 20K) (b), and the SumHer method (N = 434K) (restricted to the 24 annotations included in the SumHer model) (c) vs. functional enrichment estimates of S-LDSC+LDAK with the corresponding estimand (N = 434K); S-LDSC+LDAK (S-LDSC estimand) and S-LDSC+LDAK (LDAK estimand) produced very similar results for these 28 annotations (see Supplementary Note). Results were meta-analyzed across 16 independent UK Biobank traits. For the LDAK method (b) and the SumHer method (c), we also report results for corresponding methods using a non-default flag that models SNPs in perfect LD differently by assigning non-zero heritability to all SNPs (LDAK-nonzeroweights and SumHer-nonzeroweights, respectively). In each case we report the concordance correlation coefficient (ρ c) with S-LDSC+LDAK. Dashed grey lines represent y = x. Error bars represent 95% confidence intervals for annotations for which the estimated enrichment is significantly different (P < 0.05; two-sided z test) between the two methods. Further details and numerical results are provided in the Supplementary Note.

References

    1. Speed D, Hemani G, Johnson MR & Balding DJ Improved Heritability Estimation from Genome-wide SNPs. Am. J. Hum. Genet 91, 1011–1021 (2012). - PMC - PubMed
    1. Yang J et al. Genetic variance estimation with imputed variants finds negligible missing heritability for human height and body mass index. Nat. Genet 47, 1114–1120 (2015). - PMC - PubMed
    1. Speed D et al. Reevaluation of SNP heritability in complex human traits. Nat. Genet 49, 986–992 (2017). - PMC - PubMed
    1. Gazal S et al. Linkage disequilibrium-dependent architecture of human complex traits shows action of negative selection. Nat. Genet 49, 1421–1427 (2017). - PMC - PubMed
    1. Speed D & Balding DJ SumHer better estimates the SNP heritability of complex traits from summary statistics. Nat. Genet 51, 277–284 (2019). - PMC - PubMed

Publication types

MeSH terms

Grants and funding

LinkOut - more resources