Post by jpluta26 on Mar 28, 2019 2:36:30 GMT
Hello. I am working on a heritability analysis in a case control study. This has been well documented in the literature so I know the range of heritability I should expect. For heritability of common snps, around 37%, and for heritability of risk snps, around 9%. My estimates are proportional, but considerably lower than what I would expect. If I use the prevalence in the literature (0.005), I get snp heritability of 0.18966 and risk snp heritability of 0.0369. I'm not sure what is causing the difference. I am using a set of 10608 independent subjects, roughly equally split case/controls. There are around 7 million snps. the only covariate is a categorical variable for site. I have X chromosome data, so the female subjects are dropped, which is why n=10093.
I also notice that my estimates of risk snp heritability are proportionately lower than the literature, which is surprising because I am adding new snps. I would appreciate any insight into what could be causing my estimates to be lower.
The output is:
for common snps:
Source Variance SE
V(G) 0.062219 0.006063
V(e) 0.095022 0.005804
Vp 0.157241 0.002243
V(G)/Vp 0.395690 0.037269
The estimate of variance explained on the observed scale is transformed to that on the underlyi
ng scale:
(Proportion of cases in the sample = 0.555038; User-specified disease prevalence = 0.005000)
V(G)/Vp_L 0.189662 0.017864
logL 4299.379
logL0 4232.925
LRT 132.907
df 1
Pval 0
n 10093
for the risk snps:
Source Variance SE
V(G) 0.011949 0.002551
V(e) 0.143090 0.002021
Vp 0.155039 0.003249
V(G)/Vp 0.077074 0.015219
The estimate of variance explained on the observed scale is transformed to that on the underlyi
ng scale:
(Proportion of cases in the sample = 0.555038; User-specified disease prevalence = 0.005000)
V(G)/Vp_L 0.036943 0.007295
logL 4643.213
logL0 4232.925
LRT 820.575
df 1
Pval 0
n 10093
I also notice that my estimates of risk snp heritability are proportionately lower than the literature, which is surprising because I am adding new snps. I would appreciate any insight into what could be causing my estimates to be lower.
The output is:
for common snps:
Source Variance SE
V(G) 0.062219 0.006063
V(e) 0.095022 0.005804
Vp 0.157241 0.002243
V(G)/Vp 0.395690 0.037269
The estimate of variance explained on the observed scale is transformed to that on the underlyi
ng scale:
(Proportion of cases in the sample = 0.555038; User-specified disease prevalence = 0.005000)
V(G)/Vp_L 0.189662 0.017864
logL 4299.379
logL0 4232.925
LRT 132.907
df 1
Pval 0
n 10093
for the risk snps:
Source Variance SE
V(G) 0.011949 0.002551
V(e) 0.143090 0.002021
Vp 0.155039 0.003249
V(G)/Vp 0.077074 0.015219
The estimate of variance explained on the observed scale is transformed to that on the underlyi
ng scale:
(Proportion of cases in the sample = 0.555038; User-specified disease prevalence = 0.005000)
V(G)/Vp_L 0.036943 0.007295
logL 4643.213
logL0 4232.925
LRT 820.575
df 1
Pval 0
n 10093