collectBest: given a collection of All.cis outputs (cisRun instances)...

In GGtools: software and data for analyses in genetics of gene expression

Description

given a collection of All.cis outputs (cisRun instances) compute FDRs for various filterings

Usage

collectBest(fns, 
 targetname = "harvest", 
 mafs = c(0.01, 0.02, 0.025, 0.03333, 0.05, 0.075, 0.1), 
 hidists = c(10000, 25000, 50000, 75000, 1e+05, 250000), interimSaves=FALSE)

collectFiltered( fns, targetname="harvest",
   mafs = c(.01, .02, .025, .03333, .05, .075, .1),
   hidists = c(10000, 25000, 50000, 75000, 100000, 250000),
   filterFun = cis.FDR.filter.best, filtApplier=lapply,
   interimSaves=FALSE) 

Arguments

fns	names of .rda with the cisRun outputs
targetname	basename of rda file to be emitted
mafs	lower bounds on MAF for filtering
hidists	upper bounds on cis radius for filtering
filterFun	function like GGtools:::cis.FDR.filter.best
filtApplier	function like lapply
interimSaves	logical, if TRUE save list at each maf/dist transition

Details

pifdr is repeatedly used to generate conditional plugin FDR for different filtering criteria

Value

a list of lists is written to disk incrementally, as the job can be long running

Note

This is the workhorse of sensitivity analysis. Permits counting of genes with eQTL at selected FDR for various criteria on cis radius and MAF.

Examples

## Not run: 
#
# contents of fns are two chromsomes of cis runs for CEU
#
  fns = dir(system.file("rdas", package="GGtools"), full=TRUE)
  cc = collectBest(fns, mafs=c(.01, .05), hidists=c(10000, 50000))
  sapply(cc, sapply, function(x) sum(x$fdr <= 0.01))
#
# this tells us which to keep
#
  kp = cc[["0.05"]][["50000"]]
  kp = kp[kp$fdr <= 0.01,]
#
# the hits are in the table above; the following function
# retrieves the initial scores giving rise to the filtered
# hits
#
  pullHits = function(fns, atts) {
    tmp = lapply(fns, function(x) get(load(x)))
    kl = lapply(tmp, function(x) paste(names(x), x$snp, sep=":"))
    attk = paste(atts$genes, atts$bestsnp, sep=":")
    tmp = lapply(1:length(tmp), function(x) tmp[[x]][ match( attk, kl[[x]], nomatch=0 ) ])
    curans = do.call(c, lapply(tmp, as, "GRanges"))
    neword = match( attk, paste(names(curans), curans$snp, sep=":"))
    newfdr = atts$fdr[neword]
    curans$fdr = newfdr
    curans
    }
  pullHits( fns, kp )
#
#
#
#   after executing code in example for All.cis (protected by dontrun)
#   and running save(f1, file="f1.rda"), the following will work
# genewise max score
    cf1 = collectFiltered("f1.rda", mafs=.02, hidists=25000, targetname="gwise")
# SNPwise scores, all
    cf2 = collectFiltered("f1.rda", mafs=.02, hidists=25000, targetname="swise",
         filterFun = cis.FDR.filter.SNPcentric.complete )
# SNPwise scores, best per SNP when SNP is cis to multiple genes
    cf3 = collectFiltered("f1.rda", mafs=.02, hidists=25000, targetname="swise2",
         filterFun = cis.FDR.filter.SNPcentric )
  
## End(Not run)  # end dontrun

GGtools documentation built on Nov. 8, 2020, 6:32 p.m.