Learn R Programming

GGtools (version 5.8.0)

appraise: appraisal for eQTL prediction models

Description

appraisal for eQTL prediction models

Usage

appraise(dtab, discretize = TRUE, reduceToSNP = TRUE, prefix, folder = paste0(prefix, "_APPROUT"), discfmlas_in = GGtools:::.discfmlas.demo, txlist = list( distcats = function(x) { cut(x$mindist, c(-1, seq(0, 200001, 50000))) }, fdrcats = function(x) { fdrfac = cut(x$fdr, c(-.01, .05, .1, .25, .5, 1.01)) relevel(fdrfac, "(0.5,1.01]") }, mafcats = function(x) { maffac = cut(x$MAF,c(-0.01,.05, .1, .25, .51)) relevel(maffac, "(-0.01,0.05]") }, caddcats = function(x){ cut(x$PHRED, c(-.01, 5, seq(10, 30, 10 ), 60)) } ), cutts = c(-0.01,seq(0.015,.12,.015),.15), names2check= GGtools:::.standardNames, maxit=30, savePinfer=FALSE) # bindgwava( gwavadt, eqdt )

Arguments

dtab
data.table instance as created by transforming cisRun to GRanges and then to data.table, and then adding CADD PHRED scores if available. If CADD PHRED scores are not available, the default formulas should not be used.
discretize
logical telling whether binning to factors defined in txlist should be performed
reduceToSNP
logical telling whether ranges should be reduced to unique SNP and FDR recomputed
prefix
character atom used to prefix objects saved and folder for result objects
folder
folder name suffix
discfmlas_in
named list of model formulae
txlist
named list of functions that are used to bin certain quantitative features of SNP
cutts
numeric vector of thresholds for tabulation and discrete calibration
names2check
if NULL, ignored; if a character vector, function will fail unless all(names2check %in% names(dtab)
maxit
numeric passed to bigglm as control parameter for maximum number of iterations to use in modeling gwas hit probabilities
savePinfer
logical specifying whether the inferred probabilities of GWAS involvement are retained

Value

A folder is opened and objects are written representing the test set (data.table on SNPs on even chromosomes), the coefficients of predictive models built on training set (SNPs on odd chromosomes), coefficients of linear regressions of binary test outcomes for calibrating the model on test data, and ROC AUC measures.bindgwava uses simple data.table operations with match to add three columns to eqdt, gwava_tss, gwava_unmat, and gwava_regi

Details

The appraise function wraps many tasks used to appraise eQTL collections in terms of predictive capacity. Details will be provided.