all_adjusters: Combine all surrogate estimators and batch correctors into...

In elsayed-lab/hpgltools: A pile of (hopefully) useful R functions

Combine all surrogate estimators and batch correctors into one function.

Description

For a long time, I have mostly kept my surrogate estimators and batch correctors separate. However, that separation was not complete, and it really did not make sense. This function brings them together. This now contains all the logic from the freshly deprecated get_model_adjust().

Usage

all_adjusters(
  input,
  design = NULL,
  estimate_type = "sva",
  batch1 = "batch",
  batch2 = NULL,
  surrogates = "be",
  low_to_zero = FALSE,
  cpus = 4,
  na_to_zero = TRUE,
  expt_state = NULL,
  confounders = NULL,
  chosen_surrogates = NULL,
  adjust_method = "ruv",
  filter = "raw",
  thresh = 1,
  noscale = FALSE,
  prior_plots = FALSE
)

Arguments

input	Dataframe or expt or whatever as the data to analyze/modify.
design	If the data is not an expt, then put the design here.
estimate_type	Name of the estimator.
batch1	Column in the experimental design for the first known batch.
batch2	Only used by the limma method, a second batch column.
surrogates	Either a number of surrogates or a method to search for them.
low_to_zero	Move elements which are <0 to 0?
cpus	Use parallel and split intensive operations?
na_to_zero	Set any NA entries to 0?
expt_state	If this is not an expt, provide the state of the data here.
confounders	List of confounded factors for smartSVA/iSVA.
chosen_surrogates	Somewhat redundant with surrogates above, but provides a second place to enter because of the way I use ... in normalize_expt().
adjust_method	Choose the method for applying the estimates to the data.
filter	Filter the data?
thresh	If filtering, use this threshold.
noscale	If using combat, scale the data?
prior_plots	Plot the priors?

Details

This applies the methodologies very nicely explained by Jeff Leek at https://github.com/jtleek/svaseq/blob/master/recount.Rmd and attempts to use them to acquire estimates which may be applied to an experimental model by either EdgeR, DESeq2, or limma. In addition, it modifies the count tables using these estimates so that one may play with the modified counts and view the changes (with PCA or heatmaps or whatever). Finally, it prints a couple of the plots shown by Leek in his document. In other words, this is entirely derivative of someone much smarter than me.

Value

List containing surrogate estimates, new counts, the models, and some plots, as available.

See Also

[all_adjuster()] [isva] [sva] [limma::removeBatchEffect()] [corpcor] [edgeR] [RUVSeq] [SmartSVA] [variancePartition] [counts_from_surrogates()]

elsayed-lab/hpgltools documentation built on April 8, 2024, 1:30 a.m.