WISARD official site

Select O/S : [?]

Case tutorial

Heritability Estimation

This section describes about

Heritability estimation using pedigree structure
Heritability estimation using genotypes
Adjusting covariates for heritability estimation
Related options
- Set estimation algorithm threshold
- Use ML model instead of REML model
- Use spectral decomposition

WISARD can estimate a narrow-sense heritability of given phenotype from both of the pedigree structure regardless of genotype or given genotype. Since WISARD estimates the heritability using linear mixed model, it is possible to incorporate additional covariates, such as age, height in order to estimate the heritability.

Heritability estimation using pedigree structure [top]

Estimating heritability of the default phenotype with kinship coefficient matrix C:\Users\WISARD> wisard --bed test_miss0.bed --heritability --kinship --out res_heri_family

Column	Format	Modifier	Description
*poly.est.res* is...			A result of heritability estimation (table)
ESTNAME	string	NONE	Phenotype label that used for estimation
ESTMETHOD	ML/REML	NONE	The method used for estimation,
sig2	non-negative real	NONE	Variance by each sample
sig2g	non-negative real	NONE	Variance by sample relatedness
logL	real	NONE	Estimated log-likelihood from model
Var(sig2)	non-negative real	NONE	Variance of sig2
Var(sig2g)	non-negative real	NONE	Variance of sig2g
h^2	non-negative real	NONE	Estimated narrow-sense heritability(h^2)
Var(h^2)	non-negative real	NONE	Variance of h^2
Var(sig2+sig2)	non-negative real	NONE	Variance of entire sig

NOTE!

This run is only validate for family dataset!

Heritability estimation using genotypes [top]

Note that in order to perform heritability estimation using genotypes, a sufficient number of common(MAF >5%) autosomal markers are required, say at least ten thousand.

Estimating heritability of the default phenotype using Genetic Relationship Matrix (GRM) C:\Users\WISARD> wisard --bed test_miss0.bed --heritability --out res_heri_geno

Adjusting covariates for heritability estimation [top]

In order to adjust the effect of covariates, some covariates can be assigned using --sampvar and --cname option.

Estimating heritability with adjustment of two covariates C:\Users\WISARD> wisard --bed test_miss0.bed --heritability --sampvar test_miss0_phen.txt --kinship --cname age,height --out res_heri_geno_cov

Related options [top]

Set estimation algorithm threshold

In order to assure sufficient accuracy, a convergence threshold for estimating algorithm is fixed to $10^{-10}$. This threshold can be adjusted with --aithr option.

Perform heritability estimation with relaxed threshold C:\Users\WISARD> wisard --bed test_miss0.bed --heritability --aithr 1e-6 --out res_heri_geno_1e-6

Rather than relaxing the convergence threshold, make WISARD to not stop until converge is also possible with --nostop option. Note that this option should be careful to use.

Perform heritability estimation with UNLIMITED tries C:\Users\WISARD> wisard --bed test_miss0.bed --heritability --nostop --out res_heri_geno

Use ML model instead of REML model

WISARD using REML model for LMM fitting in default. It can be changed to ML model with --ml option.

Perform heritability estimation with ML model C:\Users\WISARD> wisard --bed test_miss0.bed --heritability --ml --kinship --out res_heri_geno_ml

Use spectral decomposition

Heritability estimation can be done with spectral decomposition method, instead of using Linear Mixed Model. In order to spectral decomposition in WISARD, use --specdcmp option.

NOTE!

Spectral decomposition requires more computation burden than LMM!

Perform heritability estimation with spectral decomposition C:\Users\WISARD> wisard --bed test_miss0.bed --heritability --kinship --specdcmp --out res_heri_geno_specdcmp

Edit this page

Last modified : 2017-09-15 20:31:30