correlate {granulator} | R Documentation |
correlate
computes Pearson correlations between
estimated cell type proportions generated by different methods.
correlate(deconvoluted, scale = TRUE)
deconvoluted |
A list: output object from |
scale |
Boolean: indicate whether the coefficients should be transformed to standard scores (default: scale = TRUE). |
correlation_analysis
is particularly useful to assess the
performance of the different methods when no ground truth is available.
If several methods agree on similar relative abundances of cell types across
samples, the results are more likely to reflect true differences in cell
type composition.
Returns a list encompassing two data frames:
the pearson correlation of coefficients with all other coefficients
summay: summary statistics of all-to-all correlation of coefficients by cell type
rank: ranking of deconvolution alghoritms by highest all-to-all correlation of coefficients
rank: ranking of deconvolution alghoritms by highest average regression all-to-all correlation of coefficients
combinations: combination of methods and signatures tested
Vincent Kuettel, Sabina Pfister
# load data load_ABIS() # deconvolute decon <- deconvolute(m = bulkRNAseq_ABIS, sigMatrix = sigMatrix_ABIS_S0) # correlate correl <- correlate(deconvoluted = decon)