storeToStats: extract a vector from store results as ff (out of memory...

In gQTLstats: gQTLstats: computationally efficient analysis for eQTL and allied studies

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

Details

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

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

Note

uses ffbase:::c.ff explicitly to concatenate outputs; there is no guarantee of order among elements

Examples

## Not run: 
 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)
  }

## End(Not run)

gQTLstats documentation built on Nov. 8, 2020, 7:53 p.m.