methylation_subtype_classfication: Classify subtypes by methylation data

In eilslabs/epic: Integrative Analysis and Visualization of Epigenomic Sequencing Data

Description

Classify subtypes by methylation data

Usage

methylation_subtype_classfication(gr, n_class, pct_cutoff = 1, corr_cutoff = 0.5,
    k, ha = NULL, cgi = NULL, shore = NULL)

Arguments

gr	a GRanges which contains mean methylation, should be generated by get_mean_methylation_in_genomic_features
n_class	number of classes expected
pct_cutoff	percent of most variable rows
corr_cutoff	cutoff for absolute correlation
k	number of correlated windows
ha	additional annotation
cgi	a GRanges object which contains CpG islands
shore	a GRanges object which contains CpG shores

Details

For the subtype classification which is based on clustering, if there are clear subtypes, it is expected that there must be a group of rows that show high correlation to each other. Based on this correlation feature, we select rows that under cutoff of corr_cutoff, each row should correlate to at least other k rows. On the second hand, since difference between subtypes are not in an identical position, we first separate samples into two groups based on consensus clustering, then, for the subgroup which contains more samples, we separate them again into two subgroups. We apply it repeatedly until there are n_class subtypes. On every step of clustering, we select rows based on the correlation criterion and the final rows are union of rows in all iterations.

CpG islands and shores will be added as row annotations to the heatmap.

Value

A vector with predicted classification of samples

Author(s)

Zuguang Gu <[e-mail]>

Examples

# There is no example
NULL

eilslabs/epic documentation built on May 16, 2019, 1:24 a.m.