evaluatePerformanceSignatureSimilarityList: evaluate performance signature similarity list

In Nik-Zainal-Group/signature.tools.lib: signature.tools.lib

evaluate performance signature similarity list

Description

Compare a signatures matrix of true signatures with a list of matrices of estimated signatures, and compute a match between the signatures and their cosine similarity. For each matrix of estimated signatures, a match implies an optimal assignment that maximises the cosine similarity while allowing each true_signature to match at most one estimated signature and viceversa. If the number of signatures differ, the least similar signatures are removed until the number of signatures is the same Results are plotted to a file if outfile is specified.

Usage

evaluatePerformanceSignatureSimilarityList(
  true_signatures,
  estimated_signatures_list,
  true_exposures = NULL,
  outfile = NULL
)

Arguments

true_signatures	matrix of true signatures, for example obtained using simulations, with mutational signatures names as colnames and channels as rownames
estimated_signatures_list	list of matrices of estimated signatures, for example obtained using a signature extraction on a simulated dataset, with mutational signatures names as colnames and channels as rownames. Use a named list, so that the names can be used in he data structure returned and plots.
true_exposures	matrix of true exposures, for example obtained using simulations, with mutational signatures names as colnames and sample names as rownames. This is optional, and it useful to show which of the true_signature are actually present in the dataset.
outfile	file name for plotting, please use .pdf file name extension. Can be omitted.

Value

cosine similarity and signatures match between true and estimated signature names

Nik-Zainal-Group/signature.tools.lib documentation built on April 13, 2025, 5:50 p.m.