calculate.post.probs: Calculate posterior probability of being a driver

In kdkorthauer/MADGiC: Fits an empirical Bayesian hierarchical model to obtain posterior probabilities that each gene is a driver.

Description

This function calculates the posterior probability for every gene being a driver

Usage

  calculate.post.probs(bij, sample.name, nonsil.mut.type,
    gid, exome.SIFT, f0, f1, alpha = 0.2, beta = 6, p0)

Arguments

bij	list object with one entry per gene, where each entry is a numeric vector containing 1-the probability of a mutation in each sample under the background mutation model
nonsil.mut.type	table of mutations, object obtained from mut.type.converter
f0	numeric vector containing the estimated null density of FI scores obtained from locfdr
f1	numeric vector containing the estimated non-null density of FI scores obtained from locfdr
alpha	numeric value of first shape parameter of the prior Beta distribution on the probability of mutation for driver genes. Default value of 0.2 is chosen as a compromise between a cancer type with a relatively low mutation rate (Ovarian cancer, fitted value from COSMIC of 0.15) and one with a comparatively high mutation rate (Squamous cell lung, fitted value from COSMIC of 0.27), but results are robust to changes in this parameter. Note that intuitively (and empirically), a higher mutation rate overall leads to a higher driver mutation rate overall - and thus less mass is concentrated in the left tail of the distribution.
beta	numeric value of second shape parameter of the prior Beta distribution on the probability of mutation for driver genes. Default value of 6 is chosen as a compromise between a cancer type with a relatively low mutation rate (Ovarian cancer, fitted value from COSMIC of 6.6) and one with a comparatively high mutation rate (Squamous cell lung, fitted value from COSMIC of 5.83), but results are robust to changes in this parameter. Note that intuitively (and empirically), a higher mutation rate overall leads to a higher driver mutation rate overall - and thus less mass is concentrated in the left tail of the distribution.
exome.SIFT	list object with one item per chromosome where each item contains matrix with one row per coding base pair and 7 columns: position, nucleotide, CpG context, FI score for mutation to "A", FI score for mutation to "C", FI score for mutation to "G", and FI score for mutation to "T".
p0	vector of prior probabilities for each gene being a driver. This is precomputed from COSMIC and loaded into package. See paper for details.
sample.name	a character vector containing the names of all samples
gid	a character vector containing the Ensembl gene ids of all genes

Value

named vector of posterior probabilities that each gene is a driver.

Note

This internal function is not intended to be called by the user.

kdkorthauer/MADGiC documentation built on June 13, 2020, 1:35 p.m.