bandle-predict: Make predictions from a bandle analysis

In ococrook/bandle: An R package for the Bayesian analysis of differential subcellular localisation experiments

Make predictions from a bandle analysis

Description

Make predictions from a bandle analysis

Usage

bandlePredict(objectCond1, objectCond2, params, fcol = "markers")

Arguments

objectCond1	A list of instances of class MSnbase::MSnSets where each is an experimental replicate for the first condition, usually a control
objectCond2	A list of instance of class MSnbase::MSnSets where each is an experimental replicate for the second condition, usually a treatment
params	An instance of class bandleParams, as generated by bandle().
fcol	A feature column indicating the markers. Defaults to "markers"

Value

bandlePredict returns an instance of class MSnbase::MSnSet containing the localisation predictions as a new bandle.allocation feature variable. The allocation probability is encoded as bandle.probability (corresponding to the mean of the distribution probability). In addition the upper and lower quantiles of the allocation probability distribution are available as bandle.probability.lowerquantile and bandle.probability.upperquantile feature variables. The Shannon entropy is available in the bandle.mean.shannon feature variable, measuring the uncertainty in the allocations (a high value representing high uncertainty; the highest value is the natural logarithm of the number of classes). An additional variable indicating the differential localization probability is also added as bandle.differential.localisation

Examples

library(pRolocdata)
data("tan2009r1")
set.seed(1)
tansim <- sim_dynamic(object = tan2009r1, 
                    numRep = 6L,
                   numDyn = 100L)
gpParams <- lapply(tansim$lopitrep, function(x) 
fitGPmaternPC(x, hyppar = matrix(c(0.5, 1, 100), nrow = 1)))
d1 <- tansim$lopitrep
control1 <- d1[1:3]
treatment1 <- d1[4:6]
mcmc1 <- bandle(objectCond1 = control1, objectCond2 = treatment1, gpParams = gpParams,
                                     fcol = "markers", numIter = 5L, burnin = 1L, thin = 2L,
                                     numChains = 1, BPPARAM = SerialParam(RNGseed = 1))
mcmc1 <- bandleProcess(mcmc1)
out <- bandlePredict(objectCond1 = control1, objectCond2 = treatment1, params = mcmc1)

ococrook/bandle documentation built on Nov. 4, 2024, 12:27 a.m.