reCluster: re-cluster clustering by 'kmeans'

In raim/segmenTools: Tools for Functional Genome Exploration

re-cluster clustering by kmeans

Description

Use cluster centers from an initial clustering to initialize kmeans. This is still experimental, and used to re-associated data rows to cluster centers from a best clustering found by flowclusterTimeseries. While the latter clustering works best to extract specific time-courses from the data set, it often comes with a high fraction of badly associated individual data sets. Re-clustering with kmeans seems to clean this up, e.g., the phase distributions of re-clustered clusterings are often tighter. TODO: allow to generate cluster centers from novel data, to account for different/more data then during clustering!

Usage

reCluster(tset, cset, k, select = TRUE, ...)

Arguments

tset	the ‘timeseries’ object from segmenTier's processTimeseries used for initial clustering
cset	the ‘clustering’ object from segmenTier's flowclusterTimeseries
k	colum name or index of the clustering that should be re-clustered; defaults to the pre-selected clustering if missing
select	use the re-clustered clustering as the new pre-selected clustering
...	parameters to kmeans

raim/segmenTools documentation built on May 5, 2024, 11:30 a.m.