Skip to contents

Prediction of exposures in new samples using pre-existing signatures

Source: R/discovery_prediction.R
predict_exposure.Rd

Exposures for samples will be predicted using an existing set of signatures stored in a result_model object. Algorithms available for prediction include a modify version of "lda", and "decompTumor2Sig".

Usage

predict_exposure(
  musica,
  modality,
  signature_res,
  algorithm = c("lda", "decompTumor2Sig"),
  result_name = "result",
  model_id = NULL,
  signatures_to_use = seq_len(ncol(signatures(signature_res))),
  verbose = FALSE,
  make_copy = FALSE,
  table_name = NULL
)

Arguments

musica

A musica object.

modality

Modality for posterior prediction. Must match the table type used to generate the prediction signatures

signature_res

Signatures used to predict exposures for the samples musica object. Existing signatures need to stored in a result_model object.

algorithm

Algorithm to use for prediction of exposures. One of "lda" or "decompTumor2Sig".

result_name

Name for result_list entry to save the results to. Default "result".

model_id

Identifier for the result. If NULL, will be automatically set to the algorithm and number of signatures. Default NULL.

signatures_to_use

Which signatures in the signature_res result object to use. Default is to use all signatures.

verbose

If TRUE, progress will be printing. Only used if algorithm = "lda". Default FALSE.

make_copy

If FALSE, the inputted musica object is updated and nothing is returned. If TRUE, a new musica object is created and returned. Default FALSE.

table_name

Use modality instead

Value

Returns nothing or a new musica object, depending on the make_copy parameter.

Examples

data(musica)
data(cosmic_v2_sigs)
g <- select_genome("19")
build_standard_table(musica, g, "SBS96", overwrite = TRUE)
#> Building count table from SBS with SBS96 schema
#> Warning: Overwriting counts table: SBS96
result <- predict_exposure(
  musica = musica, modality = "SBS96",
  signature_res = cosmic_v2_sigs, algorithm = "lda"
)

# Predict using LDA-like algorithm with seed set to 1
set.seed(1)
predict_exposure(
  musica = musica, modality = "SBS96",
  signature_res = cosmic_v2_sigs, algorithm = "lda"
)
#> model_id lda30_exp_pred already exists. model_id updated to lda30_exp_pred.1