ccr.ROC_Curve: Classification performances of reference sets of genes (or...

In francescojm/CRISPRcleanR: Unsupervised correction of gene independent cell responses to CRISPR-cas9 targeting

Classification performances of reference sets of genes (or sgRNAs) based on depletion log fold-changes

Description

This functions computes Specificity/Sensitivity (or TNR/TPR, or ROC) curve, area under the ROC curve and (optionally) Recall (i.e. TPR) at fixed false discovery rate (computed as 1 - Precision (or Positive Predicted Value)) and corresponding log fold change threshold) when classifying reference sets of genes (or sgRNAs) based on their depletion log fold-changes

Usage

ccr.ROC_Curve(FCsprofile,
                         positives,
                         negatives,
                         display = TRUE,
                         FDRth = NULL,
                         expName = NULL)

Arguments

FCsprofile	A numerical vector containing gene average depletion log fold changes (or sgRNAs' depletion log fold changes) with names corresponding to HGNC symbols (or sgRNAs' identifiers).
positives	A vector of strings containing a reference set of positive cases: HGNC symbols of essential genes or identifiers of their targeting sgRNAs. This must be a subset of FCsprofile names, disjointed from negatives.
negatives	A vector of strings containing a reference set of negative cases: HGNC symbols of essential genes or identifiers of their targeting sgRNAs. This must be a subset of FCsprofile names, disjointed from positives.
display	A logical parameter specifying if a plot containing the computed ROC curve with ROC indicators should be plotted (default = TRUE).
FDRth	If different from NULL, will be a numerical value >=0 and <=1 specifying the false discovery rate threshold at which fixed recall will be computed. In this case, if the display parameter is TRUE, an orizontal dashed line will be added to the plot at the resulting recall and its value will be visualised in the legend.
expName	If different from NULL and display parameter is TRUE this parameter should be a string specifying the title of the plot with the computed ROC curve.

Value

A list containint three numerical variable AUC, Recall, and sigthreshold indicating the area under ROC curve and (if FDRth is not NULL) the recall at the specifying false discovery rate and the corresponding log fold change threshold (both equal to NULL, if FDRth is NULL), respectively.

Author(s)

Francesco Iorio ([e-mail])

See Also

BAGEL_essential, BAGEL_nonEssential,

ccr.genes2sgRNAs, ccr.VisDepAndSig,

ccr.PrRc_Curve

Examples

## Not run: 
## loading corrected sgRNAs log fold-changes and segment annotations for an example
## cell line (EPLC-272H)
data(EPLC.272HcorrectedFCs)

## loading reference sets of essential and non-essential genes
data(BAGEL_essential)
data(BAGEL_nonEssential)

## loading library annotation
data(KY_Library_v1.0)

## storing sgRNA log fold-changes in a named vector
FCs<-EPLC.272HcorrectedFCs$corrected_logFCs$avgFC
names(FCs)<-rownames(EPLC.272HcorrectedFCs$corrected_logFCs)

## deriving sgRNAs targeting essential and non-essential genes (respectively)
BAGEL_essential_sgRNAs<-ccr.genes2sgRNAs(KY_Library_v1.0,BAGEL_essential)
BAGEL_nonEssential_sgRNAs<-ccr.genes2sgRNAs(KY_Library_v1.0,BAGEL_nonEssential)

## computing classification performances at the sgRNA level
ccr.ROC_Curve(FCs,BAGEL_essential_sgRNAs,BAGEL_nonEssential_sgRNAs)

## computing gene level log fold-changes
geneFCs<-ccr.geneMeanFCs(FCs,KY_Library_v1.0)

## computing classification performances at the sgRNA level, with Recall at 5% FDR
ccr.ROC_Curve(geneFCs,BAGEL_essential,BAGEL_nonEssential,FDRth = 0.05)

## End(Not run)

francescojm/CRISPRcleanR documentation built on April 30, 2023, 5:41 a.m.