PAC: Calculate PAC

In dswatson/CCtestr: Rigorously test cluster stability

Description

This function calculates the proportion of ambiguous clustering (PAC) for each value of k tested via consensus clustering.

Usage

PAC(consensus_mats, pac_window = c(0.1, 0.9), plot = FALSE)

Arguments

consensus_mats	A list of consensus matrices, as created by a call to consensus, M3C, or ConsensusClusterPlus.
pac_window	Lower and upper bounds for the consensus index sub-interval over which to calculate the PAC. Must be on (0, 1).
plot	Return plot of PAC scores by k?

Details

Consensus clustering is a method for testing the stability of cluster membership under resampling (Monti et al., 2003). Senbabaoglu et al. (2014) demonstrated that traditional methods for estimating optimal cluster number fail when probes are not independent, which they rarely are in omic data. The authors propose a new measure, the proportion of ambiguous clustering (PAC), which represents the increase in the empirical CDF curve for each potential cluster number k over a user-defined sub-interval of the consensus index generated by the consensus cluster algorithm. The default settings of pac_window = c(0.1, 0.9) are taken from the original PAC paper, and generally lead to stable results.

Value

A data frame with PAC scores for each value of k in consensus_mats. If plot = TRUE, then PAC scores by k are plotted.

References

Monti, S., Tamayo, P., Mesirov, J., & Golub, T. (2003). Consensus Clustering: A Resampling-Based Method for Class Discovery and Visualization of Gene Expression Microarray Data. Machine Learning, 52: 91-118.

Senbabaoglu, Y., Michailidis, G. & Li, J.Z. (2014). Critical limitations of consensus clustering in class discovery. Scientific Reports, 4:6207.

Examples

mat <- matrix(rnorm(1000 * 12), nrow = 1000, ncol = 12)
cc <- consensus(mat)
pac <- PAC(cc, plot = TRUE)

dswatson/CCtestr documentation built on May 31, 2019, 11:50 p.m.