diffExpressRankOrder: Order Differentially Expressed Genes

Description Usage Arguments Details Value Note

Description

Orders the results of a differential expression calculation, as a combination of the reported fold change and P-value for each gene.

Usage

1
2
3
4
5
6
7
8
diffExpressRankOrder(folds, pvalues, wt.folds=1, wt.pvalues=1,
		 notDE.value=0, two.sided.P.values=TRUE)

diffExpressDistanceRankOrder(folds, pvalues, dists, wt.folds=1, wt.pvalues=1,
		 wt.dists=1, notDE.value=0, two.sided.P.values=TRUE)

diffExpressRankRankOrder(folds, pvalues, ranks, wt.folds=1, wt.pvalues=1,
		 wt.ranks=1, notDE.value=0, two.sided.P.values=TRUE)

Arguments

folds

numeric vector of fold change values. Must be expresses as log2 ratio values (no expression change equals 0)

pvalues

numeric vector of same length as folds giving the P-value associated with each fold change value

dists

numeric vector of same length as folds giving the distances associated with each fold change value, where greater distance implies greater difference in expression

wt.folds

scalar weight for the contribution of fold change rankings to the result

wt.pvalues

scalar weight for the contribution of P-value rankings to the result

wt.dists

scalar weight for the contribution of distance rankings to the result

notDE.value

scalar value of folds that corresponds to not being differentially expressed. Typically zero when comparing log2 transformed fold changes, and 1 when comparing enrichment or other ratios.

two.sided.P.values

logical specifying whether the P-values came from a 2-sided or 1-sided test

Details

Fold change is a way of describing the difference between two expression values as a ratio x1/x2 . For very low magnitudes, the fold change can be artifically high. For very high magnitudes, the P-value can be artifically small (especially in NextGen sequencing datasets). This function calcualates the rank order by each criteria separately, and then uses a weighted average of those to produce the resulting gene rank order.

Value

A vector of integer indices the same length as folds, giving the permutation order of the input from highest differentially expressed (most up-regulated) to lowest (most down-regulated).

Note

In a usual 2-sided test, both the strongly up-regulated and strongly down-regulated genes will have good P-values. The function assumes this about the given P-values. If a 1-sided test that gives poor P-values for the down-regulated genes was used, set two.sided.P.value=FALSE


robertdouglasmorrison/DuffyTools documentation built on Dec. 7, 2018, 8:02 a.m.