This function will calculate the pcs using prcomp function, and correlate categorical and numerical variables from metadata.

degCovariates(counts, metadata, fdr = 0.1, scale = FALSE, minPC = 5,
  correlation = "kendall", addCovDen = TRUE, plot = TRUE)

Arguments

counts

normalized counts matrix

metadata

data.frame with samples metadata.

fdr

numeric value to use as cutoff to determine the minimum fdr to consider significant correlations between pcs and covariates.

scale

boolean to determine wether counts matrix should be scaled for pca. default FALSE.

minPC

numeric value that will be used as cutoff to select only pcs that explain more variability than this.

correlation

character determining the method for the correlation between pcs and covariates.

addCovDen

boolean. Whether to add the covariates dendograme to the plot to see covariates relationship. It will show degCorCov() dendograme on top of the columns of the heatmap..

plot

Whether to plot or not the correlation matrix.

Value

: list: a) significantCovars, covariates with FDR below the cutoff. b) plot, heatmap of the correlation found. * means pvalue < 0.05. Only variables with FDR value lower than the cutoff are colored. c) corMatrix, correlation, p-value, FDR values for each covariate and PCA pais d) effectsSignificantcovars: that is PCs correlation between covariate and PCs, e) pcsMatrix: PCs loading for each sample

Examples

data(humanGender) library(DESeq2) idx <- c(1:10, 75:85) dse <- DESeqDataSetFromMatrix(assays(humanGender)[[1]][1:1000, idx], colData(humanGender)[idx,], design=~group) res <- degCovariates(log2(counts(dse)+0.5), colData(dse))
#> #> running pca and calculating correlations for: #> un-scaled data in pca; #> pve >= 5%; #> kendall cor
res$plot
res$scatterPlot[[1]]