mts_mixModelCluster_XPR: Cluster gene expression data with Gaussian Mixture Modelling

In MultiSEp: Predict Synthetic Lethality and Other Gene Dependency Relationships from Multiomics Data

Cluster gene expression data with Gaussian Mixture Modelling

Description

Assigns samples into clusters by unsupervised analysis of continuous values for a given gene, flexibly capturing between two to five different sub-populations of samples. This function is designed for RNA-seq data but may be adapted to other data types. Each sample (for example, a cell line) is assigned a probability of belonging to each cluster. Filtering steps include ensuring a gene is present in an adequate number of samples with an expression value above the background.

Usage

mts_mixModelCluster_XPR(
  dataMatrix,
  GeneXPRthresh = 3.321928,
  NumSampleThresh = 20,
  cores
)

Arguments

dataMatrix	A gene expression matrix in gene by sample format, where row names correspond to genes and column names correspond to sample names.
GeneXPRthresh	A gene expression value that retains genes expressed in samples above that value. Defaults to 3.321928 which corresponds to log2 transformed data. A TPM value of 10 is recommended for non-log transformed data.
NumSampleThresh	Determines how many samples a gene should be expressed in, with the expression value specified by GeneXPRthresh. Defaults to 20 samples.
cores	The number of compute cores to use, default is 1.

Details

Gaussian mixture modelling with Expectation-Maximisation (EM) discovers clusters, or modes, where cardinality is determined by Bayesian Information Criterion regularisation. The method from Lubbock et al. 2013 [Nucleic Acids Research 41, W562-W568] was adapted for application to gene dependency prediction. Contiguous clusters are ensured and the ordering of clusters is taken from the value one standard deviation below the mean rather than the mean value. For example, occasionally there is overlap between clusters determined by EM such that the samples contained in one cluster encompass the range of values within a second cluster; hence samples from the first cluster are both below the minimum value and above the maximum value for any sample in the second cluster. Downstream analysis based upon the clusters, for example construction of contingency tables, assumes contiguous clusters and this property is ensured by assigning cluster boundaries with the average of minimum and maximum values from neighbouring clusters.

Value

A list object for each gene (or molecular species e.g. proteoform) in the dataMatrix object

Samples	Sample identifiers.
Values	The analysed value for each sample, for example from the input log2 gene expression value.
Cluster_Assignment	The assigned cluster value for each sample.

References

Lubbock et al. (2013) TMA Navigator: network inference, patient stratification and survival analysis with tissue microarray data [Nucleic Acids Research 41, W562-W568] \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1093/nar/gkt529")}

See Also

mts_patternDetection, mts_formatMatrix

Examples

data("depMapXPR_subset")
mixModelClusters <- mts_mixModelCluster_XPR(dataMatrix = depMapXPR_subset[1,])

MultiSEp documentation built on Aug. 27, 2026, 5:07 p.m.