BaySIC (version 1.0)

baysic.test: BaySIC Evaluation of SMGs

Description

Evaluates genes for SMGs using Bayesian posterior predictive methods

Usage

baysic.test(dat.out, fit.out, fdr.level = 0.15, fuzzy.cnt = 10000, r = NULL,subtype = NULL, PB.approx = FALSE)

Arguments

dat.out

output from baysic.data

fit.out

output from baysic.fit which utilized dat.out

fdr.level

numeric ($\in(0,1)$) defining FDR level for multiple assessment passed to fuzzy.FDR.approx. Defaults to 0.15

fuzzy.cnt

number of Monte Carlo iterations to use in approximating fuzzy FDR values passed to fuzzy.FDR.approx. Defaults to 10000.

r

Optional number of MCMC draws to thin to for Monte Carlo integration, such that r

subtype

Optional $N_s\times 2$ dataframe that defines membership of cancer subtype(s), where $N_s\leq N$. The first column of subtype should consist of subject ids (same as in dat) and the second the corresponding subtype membership. When subtype is provided, baysic.test will also generate analysis results for subtype-specific analyses.

PB.approx

logical; if TRUE, the Refined Normal Approximation (RNA) of the Poisson-Binomial distribution is used when ref.dat is a list. Defaults to FALSE.

Value

list object with the following components:

test.res

a matrix with $G$ rows containing the SMG analysis results from BaySIC. This includes the gene, the posterior predictive p-values, and fuzzy rejection probabilities under FDR level fdr.level. It will also contain results for any subtype analyses if subtype is specified.

fdr.level

value of fdr.level used

fuzzy.cnt

value of fuzzy.cnt used

subtype

value of subtype, if supplied

Details

When is.list{ref.dat} is TRUE, BaySIC evaluates whether or not a gene is an SMG using the Poisson-Binomial rather than the traditional binomial distribution. This accomodates subject-specific mutation rates given varying sequence content. When $N$ is relatively large (e.g., $N\geq50$) it is recommended that optional arguments r and PB.approx be considered to alleviate computational burden.

Examples

## Not run: 
# data(example.dat)
# data(ccds.19)
# baysic.dat.ex<-baysic.data(example.dat,ccds.19)
# snv.cat.ex<-list()
# snv.cat.ex[[1]]<-grep("[^T]C[^G]",colnames(ccds.19)[-c(1:2)])
# snv.cat.ex[[2]]<-unique(c(grep("TC.",colnames(ccds.19)[-c(1:2)]),grep(".CG",colnames(ccds.19)[-c(1:2)])))
# snv.cat.ex[[3]]<-grep(".T.",colnames(ccds.19)[-c(1:2)])
# baysic.fit.ex<-baysic.fit(baysic.dat.ex,snv.cat.ex)
# baysic.test.ex<-baysic.test(baysic.dat.ex,baysic.fit.ex)
# ## End(Not run)