Post by george on Mar 16, 2016 3:56:46 GMT
Hi,
I’m trying to use the software GCTA to estimate heritability of some traits in diabetic patients. However, I have some questions about the use of this tool.
The GWA data includes 5700 samples and ~1.2 million SNPs from 22 autosomes.
1, The variance explained by all SNPs were greatly varied with different sample size for a same trait, then is there any method to evaluate the power of sample size when estimate the heritability? For example, when I only used the Phase 1 data(~2800 subjects), the estimated heritability of eGFR (estimated Glomerular Filtration Rate, a quantitative trait) is 10%, but it increases to 20% when used all samples (5700 subjects). For the trait of height, it’s 41% for Phase 1 data and 30% for the whole data. Is it reasonable? What’s the sample size to obtain a reliable heritability? The commands I used is as follows:
gcta64 --bfile TBGWA_Clean_Ch1-22 --autosome --make-grm --out TBGWA_Clean_Ch1-22.grm
gcta64 --reml --grm TBGWA_Clean_Ch1-22.grm --pheno complication2014_5700.pheno --mpheno 1 –out reml_height.nonadj
2, For some binary outcomes(case-control study), such as stroke and cardiovascular disease, the genetic variance is 0, does it mean the genetic variants not contribute (or very very little) to the variance of phenotype? If it does, it seems unreasonable, because both stroke and CVD are affected by genetic variants. The result is as follows:
Source Variance SE
V(G) 0.000000 0.014314
V(e) 0.228220 0.014912
Vp 0.228220 0.004353
V(G)/Vp 0.000001 0.062719
logL 1307.662
logL0 1307.662
LRT 0.000
df 1
Pval 0.5
n 5499
3, As I used the online imputation server to perform imputation (https://imputationserver.sph.umich.edu/start.html). The format of imputed data is VCF(variant call format), Can I input such data to perform GCTA analysis? From the introduction, it seems the GCTA only accepts the imputed data from MACH (.mldose file). Is there any method to transform the VCF file to .mldose file?
I’m trying to use the software GCTA to estimate heritability of some traits in diabetic patients. However, I have some questions about the use of this tool.
The GWA data includes 5700 samples and ~1.2 million SNPs from 22 autosomes.
1, The variance explained by all SNPs were greatly varied with different sample size for a same trait, then is there any method to evaluate the power of sample size when estimate the heritability? For example, when I only used the Phase 1 data(~2800 subjects), the estimated heritability of eGFR (estimated Glomerular Filtration Rate, a quantitative trait) is 10%, but it increases to 20% when used all samples (5700 subjects). For the trait of height, it’s 41% for Phase 1 data and 30% for the whole data. Is it reasonable? What’s the sample size to obtain a reliable heritability? The commands I used is as follows:
gcta64 --bfile TBGWA_Clean_Ch1-22 --autosome --make-grm --out TBGWA_Clean_Ch1-22.grm
gcta64 --reml --grm TBGWA_Clean_Ch1-22.grm --pheno complication2014_5700.pheno --mpheno 1 –out reml_height.nonadj
2, For some binary outcomes(case-control study), such as stroke and cardiovascular disease, the genetic variance is 0, does it mean the genetic variants not contribute (or very very little) to the variance of phenotype? If it does, it seems unreasonable, because both stroke and CVD are affected by genetic variants. The result is as follows:
Source Variance SE
V(G) 0.000000 0.014314
V(e) 0.228220 0.014912
Vp 0.228220 0.004353
V(G)/Vp 0.000001 0.062719
logL 1307.662
logL0 1307.662
LRT 0.000
df 1
Pval 0.5
n 5499
3, As I used the online imputation server to perform imputation (https://imputationserver.sph.umich.edu/start.html). The format of imputed data is VCF(variant call format), Can I input such data to perform GCTA analysis? From the introduction, it seems the GCTA only accepts the imputed data from MACH (.mldose file). Is there any method to transform the VCF file to .mldose file?
