fast_iasva: A function for fast IA-SVA
In UcarLab/iasva: Iteratively Adjusted Surrogate Variable Analysis

Description Usage Arguments Value Examples

View source: R/fast_iasva.R

The iterative procedure of fast IA-SVA is implemented in this function (fast_iasva). fast_iasva() iteratively identifies a hidden factor for unwanted variation while accounting for all known factors, and computes its contribution (i.e., the percentage of unmodeled variation explained by the hidden factor) on the unmodeled variation in the data. If the contribution is greater than a user-defined cutoff (pct.cutoff, default = 1 the next iteration to find further hidden factors.

fast_iasva(
  Y,
  X,
  intercept = TRUE,
  num.sv = NULL,
  pct.cutoff = 1,
  num.tsv = NULL,
  tol = 1e-10,
  verbose = FALSE
)

`Y`	A SummarizedExperiment or SingleCellExperiment class containing read counts where rows represent genes and columns represent samples.
`X`	A design matrix of known variables (e.g., patient ID, gender).
`intercept`	If intercept = FALSE, the linear intercept is not included in the model.
`num.sv`	number of surrogate variables to estimate.
`pct.cutoff`	percetage threshold for SV retention. IA-SVA computes the percentage of unmodeled variance explained by the putative hidden factor and compare it with the user-defined threshold. If the percentage is greater than the threshold, SV is retained.
`num.tsv`	num of top singular values to be used in computing the percentage of unmodeled variation explained by the putative hidden factor. If num.tsv = NULL, all singular values are used.
`tol`	stopping tolerance for the augmented implicitly restarted Lanczos bidiagonalization algorithm
`verbose`	If verbose = TRUE, the function outputs detailed messages.

sv matrix of estimated surrogate variables, one column for each surrogate variable.

pct vector of percentages of unmodeled variance explained by each surrogate variable, one value for each surrogate variable.

n.sv number of obtained surrogate variables.

counts_file <- system.file("extdata", "iasva_counts_test.Rds",
 package = "iasva")
counts <- readRDS(counts_file)
anns_file <- system.file("extdata", "iasva_anns_test.Rds",
 package = "iasva")
 anns <- readRDS(anns_file)
Geo_Lib_Size <- colSums(log(counts + 1))
Patient_ID <- anns$Patient_ID
mod <- model.matrix(~Patient_ID + Geo_Lib_Size)
summ_exp <- SummarizedExperiment::SummarizedExperiment(assays = counts)
iasva.res <- fast_iasva(summ_exp, mod[, -1], num.sv = 5)