csSAM-package: csSAM package - Cell Type-Specific Statistical Analysis of...

In shenorrLab/csSAM: csSAM - cell-specific Significance Analysis of Microarrays

Description

csSAM package - Cell Type-Specific Statistical Analysis of Microarray

Details

Tissues are often made up of multiple cell-types. Each with its own functional attributes and molecular signature. Yet, the proportions of any given cell-type in a sample can vary markedly. This results in a significant loss of sensitivity in gene expression studies and great difficulty in identifying the cellular source of any perturbations. Here we present a statistical methodology (cell-type specific Significance Analysis of Microarrays or csSAM) which, given microarray data from two groups of biological samples and the relative cell-type frequencies of each sample, estimates in a virtual manner the gene expression data for each cell-type at a group level, and uses these to identify differentially expressed genes at a cell-type specific level between groups.

The lower limit for the number of samples needed for deconvolving the cell-specific expression of N cell-types is N+1. For a single colour array - the result could be interpreted as the average expression level of a given gene in a cell-type of that group. Multiplied by the frequency of a given cell-type in an individual in the group, it is the amount contributed by that cell type to the overall measured expression on the array.

Key functions for this package

• csSamWrapper - Single wrapper function performs all functionality. csfit: For deconvolving the average cell-type specific expression for each cell-type in a given group.

• csSAMfit - For fitting csSAM model on Biobase::ExpressionSet objects with the convenience of a formula interface.

• csSAM - For calculating the contrast between every pair of cells being compared between the two groups.

• fdrCsSAM - Estimate the false discovery rate for each cell-type specific comparison.

• findSigGene - Identifies the list of differentially expressed genes in a given cell-type at a given FDR cutoff.

• plotCsSAM - Plots a fdr plot of their results.

Additional functions exists (e.g., runSAM and fdrSAM to contrast csSAM with the tissue heterogeneity ignorant SAM).

References

Shen-Orr SS, Tibshirani R, Khatri P, Bodian DL, Staedtler F, Perry NM, Hastie T, Sarwal MM, Davis MM and Butte AJ (2010). “Cell type-specific gene expression differences in complex tissues.” Nature methods, 7(4), pp. 287-9. ISSN 1548-7105, doi: 10.1038/nmeth.1439 (URL: http://doi.org/10.1038/nmeth.1439).

Examples

library("csSAM")
##
## Generate random dataset
##
set.seed(143)
k <- 5 # number of cell types 
ng <- 500 # number of genes
p <- 20 # number of samples
ndiff <- 100 # number of genes differentially expressed

# true cell-specific signatures
H1 <- matrix(rnorm(5*ng), ncol=ng)
H2 <- H1
# create differential expression for 3rd cell type
H2[3,1:ndiff] <- H2[3,1:ndiff] + 5 

# cell frequency matrix per sample
cc <- matrix(runif(p*k), ncol=k) 
cc <- t(scale(t(cc), center=FALSE, scale=rowSums(cc))) 
colnames(cc) <- paste('cellType', 1:ncol(cc), sep="")

# global expression matrix
G <- rbind(cc[1:10, ] %*% H1, cc[11:p, ] %*%H2 ) + matrix(rnorm(p*ng), ncol=ng)
# sample classes (2 groups) 
y <- gl(2, p/2)  

fileName = "Example File.pdf";


# Now run, either using the wrapper 
# NB: more permutations would be needed for real data
deconvResults = csSamWrapper(G, cc, y, nperms = 50, alternative = "two.sided"
								, standardize = TRUE
								, medianCenter = TRUE
								, fileName = fileName)

# Or by calling each function independently: 
# this is useful if you want to perform only cell-specific expression 
# without differential expression.

# run csSAM
fit <- csSAMfit(t(G) ~ y | cc, nperms = 200, verbose = TRUE, .rng = 123)
top <- csTopTable(fit, alternvative = 'two.sided')
head(top[[1]])

# run SAM
tt.sam <- runSAM(G, y)
falseDiscovRSAM <- fdrSAM(G, y, tt.sam, nperms=200, alternative = 'two.sided')

plotCsSAM(fit, falseDiscovRSAM, alternative='two.sided', fileName)

shenorrLab/csSAM documentation built on May 29, 2019, 9:23 p.m.