gQTLstats (version 1.4.2)

storeToStats: extract a vector from store results as ff (out of memory reference); support statistical reductions

Description

extract a vector from store results as ff (out of memory reference); support statistical reductions

Usage

storeToQuantiles(store, field, 
    probs=c(seq(0,.999,.001), 1-(c(1e-4,1e-5,1e-6,1e-7))),
    ids = NULL, ..., checkField = FALSE, filter=force)
storeToHist(store, getter = function(x)
             as.numeric(S4Vectors::as.matrix(mcols(x)[,
             grep("permScore", names(mcols(x)))])), breaks, ids =
             NULL, filter = force)
storeToFDR(store, xprobs = c(seq(0, 0.999, 0.001), 1 - (c(1e-04,
             1e-05, 1e-06, 1e-07))), xfield = "chisq", getter =
             function(x) as.numeric(S4Vectors::as.matrix(mcols(x)[,
             grep("permScore", names(mcols(x)))])), filter = force,
            .id4coln=1, ids=NULL)

Arguments

store

instance of ciseStore-class

field

character tag, length one, must be name of a numeric field in the result set (typically something like 'chisq' in the GRanges generated by cisAssoc)

xfield

as field, for FDR computation, see Details.

ids

job ids to be used; if NULL, process all jobs

breaks

boundaries of histogram bins

...

supplied to makeRegistry for a temporary registry: typically will be a vector of package names if additional packages are needed to process results

checkField

if TRUE steps will be taken to verify that the tag to which 'field' evaluates is present in result in the first job

probs

numeric vector of probabilities with values in [0,1]. See quantile.ff.

xprobs

percentiles of the empirical distribution of the association statistic at which FDR estimates are recorded.

getter

function of a single argument that extracts a numeric vector of association scores obtained under permutation

x

instance of FDRsupp

filter

function accepting and returning GRanges instance, executed when cisAssoc result is loaded to modify that result, defaults to no-op

.id4coln

job id to be used for initial probe to determine names of fields in mcols of all jobs

Value

• storeToQuantiles and storeToHist return objects analogous to those returned by stats::quantile and graphics::hist.

  However, it should be noted that storeToQuantiles will use the quantile.ff of ffbase. For vectors of modest length, this can disagree with results of base::quantile by a few percent.

  storeToFDR and storeToFDRByProbe return an instance of FDRsupp class

Details

uses current BatchJobs configuration to parallelize extraction; reduceResults could be used for a sequential solution

Examples

stopifnot(require(geuvStore2)) 
 require(BatchJobs)
 require(gQTLBase)
 store = makeGeuvStore2()
 library(doParallel)
 if (.Platform$OS.type == "windows") {
   registerDoSEQ()
   } else registerDoParallel(cores=max(c(detectCores()-1,1)))
 smchisq = storeToFf( store, "chisq", ids=store@validJobs[1:3]) 
 smchisq
 if (.Platform$OS.type != "windows") { # avoid timeout
  qs = storeToQuantiles( store, "chisq", ids = store@validJobs[1:5],
    probs=seq(.1,.9,.1) )
  qs
  hh = storeToHist( store, ids = store@validJobs[1:5], breaks=
    c(0,qs,1e9) )
  hh$counts
  fd = storeToFDR( store, xprobs=c(seq(.05,.95,.05),.99,.999) )
  tail(getTab(fd),4)
  sss = storeToFDRByProbe( store , xprobs=c(seq(.05,.95,.05),.99) )
  tail(getTab(sss),4)
  }