FeatureSelection: Profile Feature Selection Methods
In tallulandrews/TreeOfCells: Inferring the Tree of Cells

Description Usage Arguments Details Value Examples

Methods for performing feature selection on cell-type profiles.

	top_diff_fs(merged_profiles, ntop=3000)
	dropout_fs(merged_profiles, diff=0.75)
	distribution_olap_fs(merged_profiles, quantile=0.5)
	fold_change_fs(merged_profiles, diff_t=2, max_t=1)

`merged_profiles`	Output from: merge_profiles function.
`ntop`	number of features to select.
`diff`	difference in proportion of zeros
`quantile`	quantile threshold on overlap
`diff_t`	log2 fold-change minimum threshold.
`max_t`	lower threshold on maximum expression level.

Performs feature selection on the cell-type profiles using different methods:

top_diff_fs : select genes with the biggest difference in mean expression across cell-types. dropout_fs : select genes with significantly higher difference in dropout proportion than diff (requires unscaled matrices) distribution_olap_fs : the distributions of the cell-type with the lowest expression and highest expression must by no more than proportion than quantile. (requires unscaled matrices) fold_change_fs : finds all genes passing a log2 fold-change threshold and expressed at at least max_t level in one cell-type.

a vector of gene names of the selected features.

obj <- list(
	mus = matrix(rgamma(200, shape=2.5, scale=2), ncol=10),
	rs  = matrix(rgamma(200, shape=0.75, scale=1), ncol=10),
	ds  = matrix(runif(200), ncol=10),
	Ns  = matrix(rpois(200, lambda=30), ncol=10))
fs1 <- top_diff_fs(obj, 10)
fs2 <- dropout_fs(obj)
fs3 <- distribution_olap_fs(obj)
fs4 <- fold_change_fs(obj)