ADAM2.tradeoffEO_TPR: Calculate ADAM model threshold
In DepMap-Analytics/ADAM2: Adaptive Daisy Model

Description Usage Arguments Details Value Author(s) See Also Examples

This function finds the minimum number of cell lines in which a gene needs to be fitness in order to be called core-fitness. This is defined as the n providing the best trade-off between i) coverage of priori-known essential genes in the resulting set of predicted core-fitness genes, i.e. fitness in at least n cell lines, and ii) deviance from expectation of the number of fitness genes in n cell lines.

1
2
3

ADAM2.tradeoffEO_TPR(EO,
                    TPR,
                    test_set_name)

`EO`	Profile of empirical odds values. Computed with the `ADAM2.empiricalOdds` function.
`TPR`	Profile of True positive rates for across number of cell line. Computed with the `ADAM2.truePositiveRate` function.
`test_set_name`	Name to give to the analysis, used for plotting titles.

Compare and plot the log10 odds ratios with the true positive rates to find the cross over point where the true positive rate falls below the odds ratio.

ADAM model threshold:

point

Number of cell lines for which a gene needs to be a fitness gene in order to be predicted as core-fitness gene.

C. Pacini, E. Karakoc & F. Iorio

ADAM2.empiricalOdds, ADAM2.truePositiveRate

#load in example binary depletion matrix
data(exampleDepMat)

# Generate the profiles of number of fitness genes across number of cell lines from
# observed data and corresponding comulative sums.
pprofile<-ADAM2.panessprofile(depMat=exampleDepMat)

# Generate a set of random profiles of number of genes depleted for a number of cell lines
# and corresponding cumulative sums by perturbing observed data.
nullmodel<-ADAM2.generateNullModel(depMat=exampleDepMat,ntrials = 1000)

#load a reference set of essential genes
data(curated_BAGEL_essential)

# Calculate log10 odd ratios of observed/expected profiles of cumulative number of fitness
# genes in fixed number of cell lines.
# Observed values are from the ADAM.panessprofile function and expected are the average of
# random set from ADAM2.generateNullModel
EO<-ADAM2.empiricalOdds(observedCumSum = pprofile$CUMsums,simulatedCumSum =nullmodel$nullCumSUM )

# Calculate True positive rates for fitness genes in at least n cell lines in the observed
# dependency matrix, with positive cases from a reference set of essential genes
TPR<-ADAM2.truePositiveRate(exampleDepMat,curated_BAGEL_essential)

# Calculate minimum number of cell lines a gene needs to be a fitness gene in order to
# be considered as a core-fitness gene
crossoverpoint<-ADAM2.tradeoffEO_TPR(EO,TPR$TPR,test_set_name = 'curated BAGEL essential')