ridgesum.pipeline: Run ridgesum with standard pipeline

In SeojinHwang/scadsum: nonconvex penalty with summary statistics and a reference panel

Run ridgesum with standard pipeline

Description

The easy way to run ridgesum

Usage

ridgesum.pipeline(
  cor,
  chr = NULL,
  pos = NULL,
  snp = NULL,
  A1 = NULL,
  A2 = NULL,
  ref.bfile = NULL,
  test.bfile = NULL,
  LDblocks = NULL,
  lambda = exp(seq(log(0.001), log(0.1), length.out = 20)),
  s = c(0.2, 0.5, 0.9, 1),
  destandardize = F,
  trace = 1,
  exclude.ambiguous = TRUE,
  keep.ref = NULL,
  remove.ref = NULL,
  keep.test = NULL,
  remove.test = NULL,
  sample = NULL,
  cluster = NULL,
  max.ref.bfile.n = 20000,
  nomatch = FALSE,
  ...
)

Arguments

cor	A vector of SNP-wise correlations with phenotype derived from summary statistics
chr	Together with pos, chromosome and position for cor
pos	Together with chr, chromosome and position for cor
A1	Alternative allele (effect allele) for cor
A2	Reference allele for cor (One of A1 or A2 must be specified)
ref.bfile	bfile (PLINK binary format, without .bed) for reference panel
test.bfile	bfile for test dataset
LDblocks	Either (1) one of "EUR.hg19", "AFR.hg19", "ASN.hg19", "EUR.hg38", "AFR.hg38", "ASN.hg38", to use blocks defined by Berisa and Pickrell (2015) based on the 1000 Genome data, or (2) a vector to define LD blocks, or (3) a data.frame of regions in bed format
lambda	to pass on to ridgesum
s	A vector of s
destandardize	Should coefficients from ridgesum be destandardized using test dataset standard deviations before being returned?
trace	Controls the amount of output.
exclude.ambiguous	Should ambiguous SNPs (C/G, A/T) be excluded?
keep.ref	Participants to keep from the reference panel (see parseselect)
remove.ref	Participants to remove from the reference panel(see parseselect)
keep.test	Participants to keep from the testing dataset (see parseselect)
remove.test	Participants to remove from the testing dataset (see parseselect)
sample	Sample size of the random sample taken of ref.bfile
cluster	A cluster object from the parallel package for parallel computing
max.ref.bfile.n	The maximum sample size allowed in the reference panel
...	parameters to pass to ridgesum

Value

A ridgesum.pipeline object with the following elements

beta	A list of ridgesum coefficients: one list element for each s
test.extract	A logical vector for the SNPs in test.bfile that are used in estimation.
also.in.refpanel	A logical vector for the SNPs in test.bfile that are used in lassosum.
sumstats	A data.frame of summary statistics used in estimation.
sd	The standard deviation for the testing dataset
test.bfile	The testing dataset
keep.test	Sample to keep in the testing dataset
ref.bfile	The reference panel dataset
keep.ref	Sample to keep in the reference panel dataset
lambda, s, keep.test, destandardized	Information to pass on to validate.lassosum.pipeline or pseudovalidate.lassosum.pipeline
pgs	A matrix of polygenic scores
destandardized	Are the coefficients destandardized?
exclude.ambiguous	Were ambiguous SNPs excluded?

Note

Berisa, T. & Pickrell, J. K. Approximately independent linkage disequilibrium blocks in human populations. Bioinformatics 32, 283-285 (2015).

Examples

## Not run: 
 ### Read summary statistics file ###
 ss <- fread("./data/summarystats.txt")
 head(ss)
 
 ### Convert p-values to correlations, assuming a sample size of 60000 for the p-values ###
 cor <- p2cor(p = ss$P_val, n = 60000, sign=log(ss$OR_A1))
 
 ### Run ridgesum using standard pipeline ### 
 out <- ridgesum.pipeline(cor=cor, chr=ss$Chr, pos=ss$Position, 
                          A1=ss$A1, A2=ss$A2,
                          ref.bfile=ref.bfile, test.bfile=test.bfile, 
                          LDblocks = "EUR.hg19")

## End(Not run)

SeojinHwang/scadsum documentation built on June 30, 2023, 10:52 p.m.