TitanCNA (version 1.10.0)

computeSDbwIndex: Compute the S_Dbw Validity Index for TitanCNA model selection

Description

Compute the S_Dbw Validity Index internal cluster validation from the TitanCNA results to use for model selection.

Usage

computeSDbwIndex(x, centroid.method = "median",  				data.type = "LogRatio", S_Dbw.method = "Halkidi", 				symmetric = TRUE)

Arguments

x

Formatted TitanCNA results output from outputTitanResults. See Example.

centroid.method

median or mean method to compute cluster centroids during internal cluster validation.

data.type

Compute S_Dbw validity index based on copy number (use ‘LogRatio’) or allelic ratio (use ‘AllelicRatio’).

symmetric

TRUE if the TITAN analysis was carried out using symmetric genotypes. See loadAlleleCounts.

S_Dbw.method

Compute S_Dbw validity index using Halkidi or Tong method. See details and references.

Value

list with components:

dens.bw

density component of S_Dbw index

scat

scatter component of S_Dbw index

S_DbwIndex

Sum of dens.bw and scat.

Details

S_Dbw Validity Index is an internal clustering evaluation that is used for model selection (Halkidi et al. 2002). It attempts to choose the model that minimizes within cluster variances (scat) and maximizes density-based cluster separation (Dens). Then, S_Dbw(|c_T|x z)=Dens(|c_T|x z)+scat(|c_T|x z).

In the context of TitanCNA, if data.type=‘LogRatio’, then the S_Dbw internal data consists of copy number log ratios, and the resulting joint states of copy number (c_T, forall c_T in {0 : max.copy.number}) and clonal cluster (z) make up the clusters in the internal evaluation. If data.type=‘AllelicRatio’, then the S_Dbw internal data consists of the allelic ratios. The optimal TitanCNA run is chosen as the run with the minimum S_Dbw. If data.type=‘Both’, then the sum of the S_Dbw for ‘LogRatio’ and ‘AllelicRatio’ are added together. This helps account for both data types when choosing the optimal solution.

Note that for S_Dbw.method, the Tong method has an incorrect formulation of the scat(c) function. The function should be a weighted sum, but that is not the formulation shown in the publication. computeSDbwIndex uses (ni/N) instead of (N-ni)/N in the scat formula, where ni is the number of datapoints in cluster i and N is the total number of datapoints.

References

Halkidi, M., Batistakis, Y., and Vazirgiannis, M. (2002). Clustering validity checking methods: part ii. SIGMOD Rec., 31(3):19--27. Tong, J. and Tan, H. Clustering validity based on the improved S_Dbw* index. (2009). Journal of Electronics (China), Volume 26, Issue 2, pp 258-264.

Ha, G., Roth, A., Khattra, J., Ho, J., Yap, D., Prentice, L. M., Melnyk, N., McPherson, A., Bashashati, A., Laks, E., Biele, J., Ding, J., Le, A., Rosner, J., Shumansky, K., Marra, M. A., Huntsman, D. G., McAlpine, J. N., Aparicio, S. A. J. R., and Shah, S. P. (2014). TITAN: Inference of copy number architectures in clonal cell populations from tumour whole genome sequence data. Genome Research, 24: 1881-1893. (PMID: 25060187)

See Also

outputModelParameters, loadAlleleCounts

Examples

data(EMresults)

#### COMPUTE OPTIMAL STATE PATH USING VITERBI ####
#options(cores=1)
optimalPath <- viterbiClonalCN(data, convergeParams)

#### FORMAT RESULTS ####
results <- outputTitanResults(data, convergeParams, optimalPath, 
                              filename = NULL, posteriorProbs = FALSE)

#### COMPUTE S_Dbw Validity Index FOR MODEL SELECTION ####
s_dbw <- computeSDbwIndex(results, data.type = "LogRatio", 
					centroid.method = "median", S_Dbw.method = "Tong")