Skip to contents

Count Matrix Processing for GiottoDB Objects

Source: R/processData_libraryNorm.R
processData-dbMatrix-methods.Rd

processData contains methods for GiottoDB objects that provide various normalization and scaling operations on count matrices as implemented in Giotto.

Working methods delegate to existing implementations in Giotto:

  • Library normalization (libraryNormParam)

  • Log normalization (logNormParam)

  • osmFISH normalization (osmFISHNormParam)

  • Arcsinh transformation (arcsinhNormParam)

  • L2 normalization (l2NormParam)

  • Z-score scaling (zscoreScaleParam)

  • Default normalization (defaultNormParam) - combines library + log

  • List processing - enables composable operations

Unsupported methods:

  • TF-IDF normalization (tfidfNormParam)

  • Quantile normalization (quantileNormParam)

  • Pearson residuals normalization (pearsonResidNormParam) - requires dense intermediate matrices incompatible with the sparse backend

Additionally, calculateHVF(expression_values = "scaled") and normalizeGiotto(scale_feats = TRUE) are not supported for GiottoDB objects. Feature centering is handled implicitly during PCA via db_svd.

Usage

processData(x, param, ...)

Arguments

x

dbMatrix object

param

S4 parameter class defining the transform operation. Can be:

  • libraryNormParam - library size normalization

  • logNormParam - log transformation

  • osmFISHNormParam - osmFISH normalization

  • arcsinhNormParam - arcsinh transformation

  • l2NormParam - L2/Euclidean normalization

  • zscoreScaleParam - z-score scaling

  • defaultNormParam - default normalization (library + log)

  • list - for chained operations

  • tfidfNormParam - not supported

  • quantileNormParam - not supported

  • pearsonResidNormParam - not supported

...

additional params to pass to the underlying methods

Value

A dbMatrix object

Details

All results are lazily evaluated. Please read dbplyr::collapse.tbl_sql to compute/save results to the db.

See also

processData for the generic and other methods

processExpression for use with giotto objects

normParam, scaleParam for parameter creation

Examples

if (FALSE) { # \dontrun{
# Create a dbMatrix object
library(dbMatrix)
con <- DBI::dbConnect(duckdb::duckdb(), ":memory:")
mat <- matrix(rpois(100, 5), nrow = 10)
dbmat <- dbMatrix(mat, con = con, name = "test")

# Library normalization
lib_norm <- processData(dbmat, normParam("library"))

# Log normalization
log_norm <- processData(dbmat, normParam("log"))

# Chained operations
scaled <- processData(dbmat, list(
  normParam("library"),
  normParam("log"),
  scaleParam("zscore")
))
} # }