hipo: Heritability Informed Power Optimization (HIPO)
In gqi/hipo: Heritability Informed Power Optimization (HIPO)

Description Usage Arguments Details Value References

View source: R/hipo.R

Performs heritability informed power optimization (HIPO) to conduct powerful association testing across multiple traits.

1
2
3

hipo(sumstats, out.path, HIPOD.num = 2, ldsc.path, python.path = NULL,
  ldscore.path = file.path(ldsc.path, "eur_w_ld_chr/"), maf.thr = 0.05,
  mergeallele = TRUE, truncate = NULL)

`sumstats`	A list of K containing summary statistics for K traits. Each element is a data frame that contains at least 6 columns<ef><bc><9a>`rsid`, `A1` (effect allele), `A2`, `N` (sample size), `z` (z-statistic), `pval`. For case-control studies, `N` should be the effective sample size `(ncases*ncontrols)/(ncases+ncontrols)`. Three optional columns can be provided: `chr` (chromosome number), `bp` (base pair; physical position), `freqA1` (allele frequency of A1); if provided, SNP filtering will be applied: (1) remove MHC region (26-34Mb of chromosome 6) (2) remove variants with MAF < `maf.thr` (see below for `maf.thr`).
`out.path`	The path where the LDSC intermediate outputs are stored.
`HIPOD.num`	Number of HIPO components for which to calculate z statistics and p-values. Default 2.
`ldsc.path`	The path to LDSC software.
`python.path`	The path to Python, if you need to use a version other than the default one.
`ldscore.path`	The path containing the LDSC LD score files. Default to be `file.path(ldsc.path,"eur_w_ld_chr/")`.
`maf.thr`	MAF threshold for quality control. SNPs with MAF < `maf.thr` are removed. Default 0.05, constrained between 0 and 0.5. Only effective when `freqA1` is present in `sumstats`.
`mergeallele`	Corresponds to the `--merge-allele` flag in LDSC, indicates whether to merge alleles to HapMap 3 SNPs. Default to be `TRUE`.
`truncate`	Used only for high-dimensional phenotypes. If `NULL`, use the full `coherit.mat` and `ldscint.mat`; if a decimal, truncate the eigenvectors of ldscint.mat with eigenvalues < truncate*(max eigenvalue of ldscint.mat). Recommended 0.05.

This function fits LD score regression to estimate the genetic covariance matrix and the covariance matrix of GWAS parameter estimates. Suitable eigendecomposition is then performed to find the weights that lead to the largest average non-centrality parameter of the underlying test statistic (HIPO-D1), as well as subsequent HIPO components. Z-statistics of HIPO components are computed for association testing.

A list that contains

`sumstats.all`	A data.frame containing the merged individual trait summary statistics and the z statistics (`z.HIPODx`) and p-values (`pval.HIPODx`) of HIPO components.
`coherit.mat`	Genetic covariance matrix.
`ldscint.mat`	Matrix of LD score regression intercepts.
`eigenvalue`	Eigenvalues from HIPO eigendecomposition. Proportional to the average non-centrality parameter.
`HIPOD.mat`	A matrix of which the columns are the HIPO components.

Qi, Guanghao, and Nilanjan Chatterjee. "Heritability Informed Power Optimization (HIPO) Leads to Enhanced Detection of Genetic Associations Across Multiple Traits." bioRxiv (2017): 218404.

gqi/hipo documentation built on May 18, 2019, 8:12 p.m.