supersigs is a companion R package to a method proposed by Afsari, et al. (2021, ELife) to generate mutational signatures from single nucleotide variants in the cancer genome. Note: Package is under active development.
More details on the statistical method can be found in this paper:
You can install the development version of supersigs from github using the
install_github() function from the
# Install development version from GitHub devtools::install_github("TomasettiLab/supersigs")
In brief, the
supersigs package contains two core functions:
get_signature trains a supervised signature for a given factor (e.g. smoking).
supersig <- get_signature(dt = data, factor = "smoking", wgs = F)
predict_signature uses the trained supervised signature to obtain predicted probabilities (e.g. probability of smoker) on a new dataset.
pred <- predict_signature(object = supersig, newdata = data, factor = "smoking")