bilevelAnalysisPathway: Bi-level meta-analysis - applied to pathway analysis
In nguyentin/BLMA: BLMA: A package for bi-level meta-analysis

Description Usage Arguments Details Value Author(s) References See Also Examples

Perform a bi-level meta-analysis conjunction with Impact Analysis to integrate multiple gene expression datasets

1
2
3

bilevelAnalysisPathway(kpg, kpn, dataList, groupList, splitSize = 5,
  metaMethod = addCLT, pCutoff = 0.05, percent = 0.05, mc.cores = 1,
  nboot = 200, seed = 1)

`kpg`	list of pathway graphs as objects of type graph (e.g., `graphNEL`)
`kpn`	names of the pathways.
`dataList`	a list of datasets to be combined. Each dataset is a data frame where the rows are the gene IDs and the columns are the samples.
`groupList`	a list of vectors. Each vector represents the phenotypes of the corresponding dataset in dataList, which are either 'c' (control) or 'd' (disease).
`splitSize`	the minimum number of disease samples in each split dataset. splitSize should be at least 3. By default, splitSize=5
`metaMethod`	the method used to combine p-values. This should be one of addCLT (additive method [1]), fisherMethod (Fisher's method [5]), stoufferMethod (Stouffer's method [6]), max (maxP method [7]), or min (minP method [8])
`pCutoff`	cutoff p-value used to identify differentially expressed (DE) genes. This parameter is used only when the enrichment method is "ORA". By default, pCutoff=0.05 (five percent)
`percent`	percentage of genes with highest foldchange to be considered as differentially expressed (DE). This parameter is used when the enrichment method is "ORA". By default percent=0.05 (five percent). Please note that only genes with p-value less than pCutoff will be considered
`mc.cores`	the number of cores to be used in parallel computing. By default, mc.cores=1
`nboot`	number of bootstrap iterations. By default, nboot=200
`seed`	seed. By default, seed=1.

The bi-level framework combines the datasets at two levels: an intra-experiment analysis, and an inter-experiment analysis [1]. At the intra-level analysis, the framework splits a dataset into smaller datasets, performs pathway analysis for each split dataset using Impact Analysis [2,3], and then combines the results of these split datasets using metaMethod. At the inter-level analysis, the results obtained for individual datasets are combined using metaMethod

A data frame (rownames are geneset/pathway IDs) that consists of the following information:

Name: name/description of the corresponding pathway/geneset
Columns that include the pvalues obtained from the intra-experiment analysis of individual datasets
pBLMA: p-value obtained from the inter-experiment analysis using addCLT
rBLMA: ranking of the geneset/pathway using addCLT
pBLMA.fdr: FDR-corrected p-values

Tin Nguyen and Sorin Draghici

[1] T. Nguyen, R. Tagett, M. Donato, C. Mitrea, and S. Draghici. A novel bi-level meta-analysis approach – applied to biological pathway analysis. Bioinformatics, 32(3):409-416, 2016.

[2] A. L. Tarca, S. Draghici, P. Khatri, S. S. Hassan, P. Mittal, J.-s. Kim, C. J. Kim, J. P. Kusanovic, and R. Romero. A novel signaling pathway impact analysis. Bioinformatics, 25(1):75-82, 2009.

[3] S. Draghici, P. Khatri, A. L. Tarca, K. Amin, A. Done, C. Voichita, C. Georgescu, and R. Romero. A systems biology approach for pathway level analysis. Genome Research, 17(10):1537-1545, 2007.

[4] R. A. Fisher. Statistical methods for research workers. Oliver & Boyd, Edinburgh, 1925.

[5] S. Stouffer, E. Suchman, L. DeVinney, S. Star, and J. Williams, RM. The American Soldier: Adjustment during army life, volume 1. Princeton University Press, Princeton, 1949.

[6] L. H. C. Tippett. The methods of statistics. The Methods of Statistics, 1931.

[7] B. Wilkinson. A statistical consideration in psychological research. Psychological Bulletin, 48(2):156, 1951.

bilevelAnalysisGeneset, pe, phyper

# load KEGG pathways
x <- loadKEGGPathways()  

# load example data
dataSets <- c("GSE17054", "GSE57194", "GSE33223", "GSE42140")
data(list=dataSets, package="BLMA")
names(dataSets) <- dataSets
dataList <- lapply(dataSets, function(dataset) get(paste0("data_", dataset)))
groupList <- lapply(dataSets, function(dataset) get(paste0("group_", dataset)))

IAComb <- bilevelAnalysisPathway(x$kpg, x$kpn, dataList, groupList)
head(IAComb[, c("Name", "pBLMA", "pBLMA.fdr", "rBLMA")])