findCopyNumber: Find copy number regions using expression data in a similar...
In phenoTest: Tools to test association between gene expression and phenotype in a way that is efficient, structured, fast and scalable. We also provide tools to do GSEA (Gene set enrichment analysis) and copy number variation.

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

Given enrichment scores between two groups of samples and the chromosomical positions of those enrichment scores this function finds areas where the enrichemnt is bigger/lower than expected if the positions where assigned at random. Plots of the regions and positions of the enriched regions are provided.

findCopyNumber(x, minGenes = 15, B = 100, p.adjust.method = "BH",
pvalcutoff = 0.05, exprScorecutoff = NA, mc.cores = 1, useAllPerm = F,
genome = "hg19", chrLengths, sampleGenome = TRUE, useOneChr = FALSE,
useIntegrate = TRUE,plot=TRUE,minGenesPerChr=100)

`x`	An object of class `data.frame` with gene or probe identifiers as row names and the following columns: es (the enrichment score), chr (the chromosome where the gene or probe belong to) and pos (position in the chromosome in megabases). It can be obtained (from an epheno object) with the function getEsPositions.
`minGenes`	Minimum number of genes in a row that have to be enriched to mark the region as enriched. Has to be bigger than 2.
`B`	Number of permuations that will be computed to calculate pvalues. If `useAllPerm` is FALSE this value has to be bigger than 100. If `useAllPerm` is TRUE the computations are much more expensive, therefore it is not recommended to use a B bigger than 100.
`p.adjust.method`	P value adjustment method to be used. p.adjust.methods provides a list of available methods.
`pvalcutoff`	All genes with an adjusted p value lower than this parameter will be considered enriched.
`exprScorecutoff`	Genes with a smoothed score that is not bigger (lower if the given number is negative) than the specified value will not be considered significant.
`mc.cores`	Number of cores to be used in the computation. If `mc.cores` is bigger than 1 the `multicore` library has to be loaded.
`useAllPerm`	If FALSE for each gene only permutations of genes that are in an area with similar density (similar number of genes close to them) are used to compute pvalues. If TRUE all permutations are used for each gene. We recommend to use the option FALSE after having observed that the enrichment can depend on the number of genes that are in the area. We recommend to use the option TRUE if the positions of the enrichment score are equidistant. Take into account that this option is much slower and needs less permutations, therefore a smaller `B` is preferred. See details for more info.
`genome`	Genome that will be used to draw cytobands.
`chrLengths`	An object of class `numeric` containing chromosome names as names. This names have to be the same as the ones used in `x$chr` If missing the last position is used.
`sampleGenome`	If positions are sampled over the hole genome (across chromosomes) or within each chromosome. This is TRUE by default.
`useOneChr`	Use only one chromosome to build the distribution under the null hypothesis that genes/probes are not enriched. By default this is FALSE. The chromosome that is used is chosen as follows: after removing small chromosomes we select the one closest to the median quadratic distance to 0. Setting this parameter to TRUE decreases processing time.
`useIntegrate`	If we want to use `integrate` or `pnorm` to compute pvalues. The first does not assume any distribution for the distribution under the null hypothesis, the second assumes it is normally distributed.
`plot`	If FALSE the function will make no plots.
`minGenesPerChr`	Chromosomes with less than `minGenesPerChr` will be removed from the analysis.

Enrichemnt scores can be either log fold changes, log hazard ratios, log variabiliy ratios or any other score.

Within each chromosome a smoothed score for each gene is obtained via generalized additive models, the smoothing parameter for each chromosome being chosen via cross-validation. The obtained smoothing parameter of each chromosome will be used in permutations.

We assessed statistical significance by permuting the positions thrue the hole genome. If useAllPerm is FALSE for each gene the permutations of genes that are in an area with similar density (distance to tenth gene) are used to compute pvalues. We observed that genes with similar densities tend to have similar smoothed scores. If we set 1000 permutations (B=1000) scores are permuted thrue the hole genome 10 times (1000/100). For each smoothed scored the permutations of the 100 smoothed scores with most similar density (distance to tenth gene) are used. Therefore each smoothed score will be compared to 1000 smoothed scores obtained from permutations.

If scores are at the same distance in the genome from each other (for instance when we have a score every fixed certain bases) the option useAllPerm=TRUE is recommended. In this case every smoothed score is compared to all smoothed scores obtained via permutations. In this case having 20,000 genes and setting the paramter B=10 would mean that the scores are permuted 10 times times thrue the hole genome, obtaining 200,000 permuted smoothed scores. Each observed smoothed score will be tested against the distribution of the 200,000 permuted smoothed scores.

Only regions with as many genes as told in minGenes being statistically significant (pvalue lower than parameter pvalcutoff) after adjusting pvalues with the method specified in p.adjust.method will be selected as enriched. If exprScorecutoff is different form NA, a gene to be statistically significant will need (aditionally to the pvalue cutoff) to have a smoothed score bigger (lower if exprScorecutoff is negative) than the specified value.

Plots all chromomes and marks the enriched regions. Also returns a data.frame containing the positions of the enriched regions. This output can be passed by to the genesInArea function to obtain the names of the genes that are in each region.

Evarist Planet

getEsPositions, genesInArea

data(epheno)
phenoNames(epheno)
mypos <- getEsPositions(epheno,'Relapse')
mypos$chr <- '1' #we set all probes to chr one for illustration purposes
                 #(we want a minimum number of probes per chromosome) 
head(mypos)
set.seed(1)
regions <- findCopyNumber(mypos,B=10,plot=FALSE) 
head(regions)