estDE: Estimate Differential Expression states

In MalathiSIDona/pathDESeq: Pathway-Based Differential Expression Analysis for RNA-Seq Data

Description

This function estimates the Differential Expression (DE) states for a considered set of genes based on gene their expression data, given Poisson-Gamma-Beta Markov Random Field model parameters and known gene network information. This function uses the Gaussian quadrature method to calculate the log likelihood value for the 3-states PGBMRF model using "statmod" package in R. Finally, it estimates the DE states based on the calculated log likelihood value and sample means.In pathDESeq package this function uses as a internal function for the pgbmrfICM function.

Usage

estDE(data, m, n, PGB.parameters, MRF.parameters, neib.matrix, state, k)

Arguments

data	A data frame which consists the gene expression data. First m columns correspond to the control group and next n columns correspond to the treatment group. Each row represents the gene expression data for a particular gene.
m	Number of replicates for the control group
n	Number of replicates for the treatment group
PGB.parameters	A vector of Poisson-Gamma-Beta parameter estimates with 4 elements as Gamma Shape parameter, Gamma Rate parameter, Beta Shape1 parameter and Beta Shape2 parameter respectively.
MRF.parameters	A vector of parameters for Markov Random Field model as gamma.u, gamma.d, beta1.u, beta2.u, beta1.d and beta2.d respectively. gamma.u = an arbitrary parameter corresponding to the up regulated DE state gamma.d = an arbitrary parameter corresponding to the down regulated DE state beta1.u = an arbitrary parameter corresponding to the up regulated DE state which encourages neighbouring genes to have same DE state beta2.u = an arbitrary parameter corresponding to the up regulated DE state which discourages to have different DE states. beta1.d = an arbitrary parameter corresponding to the down regulated DE state which encourages neighbouring genes to have same DE state beta2.d = an arbitrary parameter corresponding to the down regulated DE state which discourages to have different DE states.
neib.matrix	A matrix with 0's and 1's represents the neighbourhood information for the given gene set. For a particular gene, its neighbouring genes will have value 1 and non-neighbouring genes will have value 0. The gene itself is considered as a neighbour with value 1. For a given gene list with n number of genes neib.matrix should be a n by n matrix with all 0's and 1's while 1 represents neighbours and 0 represent non-neighbours.
state	A data matrix of differential expression states for given genes. Should be in the same gene order as in the expression data and the neighbourhood matrix.
k	Number of Gaussian quadrature points generated from the Gamma distribution and Beta distribution to numerically evaluate the double integrals. The default is 40.

Details

It is required to install the "statmod" R package.

Value

This function will return a vector of DE states which represents the estimated DE states for given genes based on available gene expression data and network information.

Author(s)

Malathi S.I. Dona

References

Smyth, G., Hu, Y., Dunn, P., Phipson, B., & Chen, Y. (2015). statmod: Statistical Modeling. R package version 1.4. 22

Smyth, G. K. (1998). Polynomial approximation. In: Encyclopedia of Biostatistics, P. Armitage and T. Colton (eds.), Wiley, London, pages 3425-3429. http://www.statsci.org/smyth/pubs/EoB/bap064-.pdf

See Also

neibMat, pgbmrfICM

Examples

#load a data frame of simulated DE states for selected 1065 Reactome pathway genes
data("state1065")
#save as a matrix
state1065<-as.matrix(state1065)

#load a data frame with 1065 observations for 21 variables.
#First column consists gene names. The next 10 columns correspond to the control group and
#the last 10 columns correspond to the treatment group.
#Each row represents gene expression data for a particular gene.
#-corresponding to treatment group. Each row represents the gene expression data for a
#-particular gene.
data("data1065")
#store gene expression data
data<-data1065[,-1]

#load an example neighbourhood matrix for selected 1065 genes
data("neib1065")
neib1065<-as.matrix(neib1065)

#estimate DE states
#define PGB  parameters as 9,0.5,2 and 2.
#define MRF parameters as 1.5,1.5,6,6,5 and 5.
#take m=10,n=10.

state<-estDE(data=data, m=10, n=10, PGB.parameters=c(9,0.5,2,2),
MRF.parameters=c(1.5,1.5,6,6,5,5), neib.matrix=neib1065, state=state1065)

#obtain a frequency table for estimated DE states
table(state)

MalathiSIDona/pathDESeq documentation built on May 8, 2019, 3:37 p.m.