compute.bic: Computes BIC for a Clomial model.

In Clomial: Infers clonal composition of a tumor

Description

Computes the Bayesian Information Criterion (BIC) for a Clomial model, which might be useful to estimate the number of clones. A "significantly" smaller BIC is usually interpreted as a better fit to the data.

Usage

	compute.bic(Dc, Dt, Mu, P)

Arguments

Dt	A matrix which contains the counts of the alternative allele where rows correspond to the genomic loci, and columns correspond to the samples.
Dc	A matrix which contains the counts of the total number of mapped reads where rows correspond to the genomic loci, and columns correspond to the samples.
Mu	The matrix which models the genotypes, where rows and columns correspond to genomic loci and clones, accordingly.
P	The matrix of clonal frequency where rows and columns correspond to clones and samples, accordingly.

Details

The Bayesian Information Criterion (BIC) for a model is computed by subtracting the expected log-likelihood times 2, from the number of free parameters of the model times logarithm of the total number of observations. For a Clomial model, we have BIC = (NC+SC-S)log(sum(Dc))-2L, where L is the likelihood, N is the number of genomic loci, C is the assumed number of clones, S is the number of samples, and sum(Dc) is the total number of observed reads.

Value

A list will be made with the following entries:

bic	The BIC value.
aic	The AIC value.
obsNum	The total number of observed reades.

Note

Theoretically, a method such as the Bayesian information criterion (BIC) or the Akaike information criterion (AIC) may be applied to estimate the number of clones. However, in practice, the outcome of such approaches should be interpreted with great caution because some of the underlying assumptions of the statistical analysis may not be necessarily true for a given model. For example, while a "small" improvement in the BIC is generally considered as a sign to stop making the model more complicated, making such decisions is very objective, and requires relying on thresholds with little statistical basis.

Author(s)

Habil Zare

References

Inferring clonal composition from multiple sections of a breast cancer, Zare et al., Submitted.

See Also

Clomial

Examples

set.seed(1)
data(breastCancer)
Dc <- breastCancer$Dc
Dt <- breastCancer$Dt
bics <- c()
Clomial3 <-Clomial(Dc=Dc,Dt=Dt,maxIt=20,C=3,doParal=FALSE,binomTryNum=1)
model3 <- Clomial3$models[[1]]
bics[3] <- compute.bic(Dc=Dc,Dt=Dt, Mu=model3$Mu, P=model3$P)$bic
Clomial4 <-Clomial(Dc=Dc,Dt=Dt,maxIt=20,C=4,doParal=FALSE,binomTryNum=1)
model4 <- Clomial4$models[[1]]
bics[4] <- compute.bic(Dc=Dc,Dt=Dt, Mu=model4$Mu, P=model4$P)$bic
print(bics) ## 4 is a better estimate for the number of clones.

Clomial documentation built on Nov. 8, 2020, 8:16 p.m.