associatePosteriors: Associates posterior likelihood to generate co-expression...

In tjh48/clusterSeq: Clustering of high-throughput sequencing data by identifying co-expression patterns

Description

This function aims to find pairwise dissimilarities between genes. It does this by comparing the posterior likelihoods of patterns of differential expression for each gene, and estimating the likelihood that the two genes are not equivalently expressed.

Usage

associatePosteriors(cD, maxsize = 250000, matrixFile = NULL)

Arguments

cD	A countData object containing posterior likelihoods of differential expression for each gene.
maxsize	The maximum size (in MB) to use when partitioning the data.
matrixFile	If given, a file to write the complete (gzipped) matrix of pairwise distances between genes. Defaults to NULL.

Details

In comparing two genes, we find all patterns of expression considered in the '@groups' slot of the 'cD' (countData) object for which the expression of the two genes can be considered monotonic. We then subtract the sum the posterior likelihods of these patterns of expression from 1 to define a likelihood of dissimilarity between the two genes.

Value

A data.frame which for each gene defines its nearest neighbour of higher row index and the dissimilarity with that neighbour.

Author(s)

Thomas J. Hardcastle

See Also

makeClusters makeClustersFF kCluster

Examples

# load in analysed countData (baySeq package) object
library(baySeq)
data(cD.ratThymus, package = "clusterSeq")

# estimate likelihoods of dissimilarity on reduced set
aM <- associatePosteriors(cD.ratThymus[1:1000,])

tjh48/clusterSeq documentation built on May 31, 2019, 3:40 p.m.