CONFESS (version 1.0.2)

Fluo_CV_prep: Fluo_CV_prep

Description

It generates the data that will be used in the cross-validation analysis. Essentialy, it analyzes and stores the original (full) dataset for different reference runs, seeds, starting clusters etc. It estimates the progression path automatically that is feasible only for standard paths (path.type parameter different than 'other'). For this reason this function is useful only in these cases. If otherwise, it should be ommitted from the analysis and the user is should generate it manually, i.e. run Fluo_adjustment() - Fluo_modeling() series as many times as the cases to be studied with manual init.path input in Fluo_modeling().

Usage

Fluo_CV_prep(data, init.path = "bottom/left", path.type = c("circular", "clockwise"), BGmethod = "normexp", maxMix = 3, single.batch.analysis = 1:5, transformation = "log", prior.pi = 0.1, flex.reps = 50, flexmethod = "BIC", areacut = 0, fixClusters = 0, altFUN = "kmeans", k.max = 15, VSmethod = "DDHFmv", CPmethod = "ECP", CPgroups = 5, B.kmeans = 50, CPpvalue = 0.05, CPmingroup = 15, savePlot = getwd(), seed = NULL)

Arguments

data

List. The output of crearteFluo(), i.e. the image analysis estimates.

init.path

Character vector. It defines the starting cluster of the progression path in general terms. It can be one of "top/right", "top/left", "bottom/right" or "bottom/left" indicating the cluster of interest on the 2d scatterplot of Fluo_inspection(). Default is rep("bottom/left",2), i.e. in Fucci an EM/earlyG1 like cluster.

path.type

Character vector. A user-defined vector that characterizes the cell progression dynamics. The first element can be either "circular" or "A2Z" or "other". If "circular" the path progression is assummed to exhibit a circle-like behavior. If "A2Z" the path is assumed to have a well-defined start and a well-defined end point (e.g. a linear progression). If "other" the progression is assumed to be arbitrary without an obvious directionality. Default is "circular". The second element can be either "clockwise" or "anticlockwise" depending on how the path is expected to proceed. Default is "clockwise". If the first element is "other" the second element can be ommited.

If path.type = "other", the function does not estimate a path. The cross-validation algorithm will probably fail for this kind of path.type values because it will not be able to automatically guess the progression path. It is suggested that the user runs the cross-validation manually (each time specifying the path in Fluo_modeling()), collect the data in a list similar to the one produced here and input them into Fluo_CV_modeling() to get the results.

BGmethod

Character string. The type of image background correction to be performed. One of "normexp" or "subtract". Default is "normexp".

maxMix

Integer. The maximum number of components to fit into the mixture of regressions model. If maxMix=1 or if the the optimal number of the estimated components is 1, the model reduces to the classical 2-way ANOVA. Default is 3.

single.batch.analysis

Numeric. The baseline run(s) to perform run effect correction with flexmix. Due to iterative nature of this function it can be a series of values includying 0 (averaging of run correction estimates). Default is 1:5.

transformation

Character string. One of bc (Box-Cox), log, log10, asinh transforms applied to the data. Default is "log".

prior.pi

Float. The prior probability to accept a component. Default is 0.1.

flex.reps

Integer. The iterations of the Expectation-Maximization algorithm to estimate the flexmix model. Default is 50.

flexmethod

Character string. A method to estimate the optimal number of flexmix components. One of "BIC", "AIC", "ICL". Default is "BIC".

areacut

Integer. The "artificial" area size (BFarea^2) of the cells estimated by BF image modelling. Default is 0, implying that the area sizes to be corrected will by estimated automatically from the data (not recommended if prior knowledge exists).

fixClusters

Integer. A number that defines the number of k-mean clusters to be initially generated. If 0, the function runs GAP analysis to estimate the optimal number of clusters. Default is 0.

altFUN

Character string. A user-defined method to generate the initial clusters. It can be one of kmeans, samSpec, fmeans,fmerge or fpeaks. Default is "kmeans".

k.max

Integer. This is the maximum number of clusters that can be generated by k-means (if fixClusters = 0). Default is 15.

VSmethod

Character string. The variance stabilization transformation method to be applied to the corrected fluorescence data prior to the change point analysis. IT can be one of "log" or "DDHFmv". Default is "DDHFmv".

CPmethod

Character string. The change point method to be used. It can be one of "ECP", (non-parametric) "manualECP" (non-parametric with user-defined numner of change-points) or "PELT" (Pruned Exact Linear Time; parametric). Default is ECP.

CPgroups

Integer. The number of change-points to be kept if CPmethod = "manualECP". Default is 5.

B.kmeans

Integer. The number of bootstrap samples for the calculation of the GAP statistic. Default is 50.

CPpvalue

Float. The significance level below which we do not reject a change point. Default is 0.05.

CPmingroup

Integer. The minimum number of values for a cluster re-estimated by the change-point analysis. Default is 10.

savePlot

Character string. Directory to store the plots of the analysis of the whole data. Its value can be an existing directory or "screen" that prints the plot only on the screen. The "OFF" option is permanently used in cross-validations). Default is the current working directory, getwd().

seed

Integer. An optional seed number for the Random Number Generator. Note that this seed is a 'reference' value of the actual seed used in sampling. CONFESS is using various random sampling methods. Each method's actual seed is factor*seed. The factors vary across methods. Default is NULL.

Value

The results of Fluo_modeling() for difference reference runs (batches) are stored in different slots. An additional slot @init.path exists that stores the init.path parameter (its value to be used in the CV automatically).One can directly use the run components in Fluo_ordering() to finalize the data analysis. The main purpose of this function, though, is to prepare the data for cross-validation.

Details

The function can also be used to generate all pseudotime/clustering results up to the function of Fluo_modeling() but the starting cluster has to be defined in general terms (see init.path parameter below). For this reason, its parameters are essentially the same to the ones defined previously at the Fluo_adjustment() - Fluo_modeling() functions.

Examples

step1 <- createFluo(from.file=system.file("extdata", "Results_of_image_analysis.txt",
package = "CONFESS"),separator="_")
steps2_4 <- Fluo_CV_prep(data=step1,init.path = "bottom/left",path.type=c("circular","clockwise"),
single.batch.analysis = 5,flex.reps=5,altFUN="kmeans",VSmethod="DDHFmv",CPmethod="ECP",
B.kmeans=5,CPpvalue=0.01,savePlot="OFF")