Fig 1.
Scenario I: QQ plots for pleiotropic analysis of null data on traits from 2 independent case-control studies.
Observed(−log10p-values) are plotted on the y-axis and Expected(−log10p-values) on the x-axis. Either each study has 1, 000 unrelated cases and 1, 000 unrelated controls, or Study 1 is 4 times that of Study 2, where Study 2 has 1, 000 unrelated cases and 1, 000 unrelated controls. Type I error performance of tests of pleiotropic effect of a genetic variant on the 2 traits is based on 9.99 million null variants with genetic effects that are either {β1 = 0 = β2} or {β1 = 0, β2 = log(1.15)} or {β1 = log(1.15), β2 = 0}. The gray shaded region represents a conservative 95% confidence interval for the expected distribution of p-values. P-values ≥10-10 are shown here.
Fig 2.
Scenario II: QQ plots for pleiotropic analysis of null data on traits from 2 case-control studies with different proportions of overlapping controls.
Observed(−log10p-values) are plotted on the y-axis and Expected(−log10p-values) on the x-axis. Equal study sample size, and equal case-control size assumed in each study. Each study has 1, 000 unrelated cases and 1, 000 unrelated controls, of which either 20%, 40%, 80% or 100% of the controls are shared between the two studies. Type I error performance of tests of pleiotropic effect of a genetic variant on the 2 traits is based on 9.99 million null variants with genetic effects that are either {β1 = 0 = β2} or {β1 = 0, β2 = log(1.15)} or {β1 = log(1.15), β2 = 0}. The gray shaded region represents a conservative 95% confidence interval for the expected distribution of p-values. P-values ≥10-10 are shown here.
Fig 3.
Scenario I: Power of PLACO, maxP and naive approaches at genome-wide significance level (5 × 10−8) for varying genetic effects of traits from 2 independent case-control studies.
Sobel’s approch is excluded from this figure since it has <1% power across all scenarios. The first naive approach (‘Naive-1’) declares pleiotropic association when pTrait1<5 × 10−8 and pTrait2<5 × 10−5, while the second naive approach (‘Naive-2’) uses a more liberal criterion pTrait1<5 × 10−8 and pTrait2<5 × 10−3. Each study either has 1, 000 unrelated cases and 1, 000 unrelated controls, or Study 1 has 4 times sample size as Study 2, where Study 2 has 1, 000 unrelated cases and 1, 000 unrelated controls.
Fig 4.
Manhattan plot of the PLACO p-values of pleiotropic association of common genetic variants with outcomes (traits) T2D and PrCa.
The black horizontal dashed line corresponds to genome-wide significance level α = 5 × 10−8. The 44 loci with genome-wide significant pleiotropic lead SNP have been highlighted. A locus is defined by clumping SNPs in ±500 Kb radius around the lead SNP and with LD r2>0.2. Within each locus, if a PLACO significant SNP has genetic effects in opposite directions for T2D and PrCa, it is plotted as a solid triangle (24 such loci), else as a solid circle. Each identified pleiotropic locus is categorized (color-coded) as follows. Three loci harbor SNPs that are marginally genome-wide significant for both T2D and PrCa (single-trait p<5 × 10−8). Four loci contain SNPs that are marginally genome-wide significant for one disease, and in close proximity (i.e., in the same locus) with another SNP marginally genome-wide significant for the other disease. There are 10 loci where SNPs are marginally genome-wide significant for one disease and in close proximity with another SNP marginally suggestively significant (single-trait p<10−5) for the other disease. Two loci harbor SNPs that are marginally suggestively significant (but not genome-wide significant) for both T2D and PrCa. There is no locus that contains SNPs that are marginally suggestively significant (but not genome-wide significant) for one disease, and in close proximity with another SNP marginally suggestively significant (but not genome-wide significant) for the other disease. The rest of the 25 loci identified by PLACO contain SNPs that are not even marginally suggestively significant for either T2D or PrCa.
Table 1.
The coloc colocalization posterior probability () for the lead SNPs from each of the 43 pleiotropic loci identified by PLACO.
Table 2.
The potentially novel loci detected by PLACO and with convincing evidence ( and
) of being causal for both T2D and PrCa from colocalization analysis.
Fig 5.
Regional association plot of significant pleiotropic locus near RGS17 with annotations such as CADD scores, RegulomeDB scores, and cis eQTL association p-values from 6 tissues.
Tissues considered are whole blood from eQTLGen Consortium; and adipose, liver, muscle-skeletal, pancreas, and prostate tissues from GTEx v8.
Fig 6.
Regional association plot of significant pleiotropic locus near UBAP2 with annotations such as CADD scores, RegulomeDB scores, and cis eQTL association p-values from 6 tissues.
Tissues considered are whole blood from eQTLGen Consortium; and adipose, liver, muscle-skeletal, pancreas, and prostate tissues from GTEx v8.