Min-Max Normalization — minmax_normalization • dominatR

Scales each column of a matrix (or SummarizedExperiment assay) so that the minimum value in that column is mapped to new_min and the maximum value is mapped to new_max

Usage

minmax_normalization(
  x,
  new_min = 0,
  new_max = 1,
  assay_name = NULL,
  new_assay_name = NULL
)

Arguments

x: A numeric matrix, data.frame, or SummarizedExperiment.
new_min: The lower bound of the new range (default 0).
new_max: The upper bound of the new range (default 1).
assay_name: If x is a SummarizedExperiment, name of the assay to normalize. Defaults to the first assay if none is specified.
new_assay_name: If x is a SummarizedExperiment, name of a new assay to store the normalized data. If NULL, overwrites the assay specified by assay_name.

Value

If x is a data.frame or matrix, returns a matrix of column-wise scaled values (same dimensions as x).
If x is a SummarizedExperiment, returns the same SummarizedExperiment object with the chosen or new assay replaced by the scaled values.

Examples

library(SummarizedExperiment)
library(airway)
data('airway')

se <- airway

# Only use a random subset of 1000 rows
set.seed(123)
idx <- sample(seq_len(nrow(se)), size = min(1000, nrow(se)))
se <- se[idx, ]

# -------------------------------
# 1) Using a data.frame
# -------------------------------

df <- assay(se)

df1 <- minmax_normalization(df)

apply(df1, 2, range)
#>      SRR1039508 SRR1039509 SRR1039512 SRR1039513 SRR1039516 SRR1039517
#> [1,]          0          0          0          0          0          0
#> [2,]          1          1          1          1          1          1
#>      SRR1039520 SRR1039521
#> [1,]          0          0
#> [2,]          1          1

## Using a new range
df1 <- minmax_normalization(df, new_min = 5, new_max = 10)

apply(df1, 2, range)
#>      SRR1039508 SRR1039509 SRR1039512 SRR1039513 SRR1039516 SRR1039517
#> [1,]          5          5          5          5          5          5
#> [2,]         10         10         10         10         10         10
#>      SRR1039520 SRR1039521
#> [1,]          5          5
#> [2,]         10         10

# -------------------------------
# 2) Using a SummarizedExperiment
# -------------------------------

# If now new_assay_name is provided, then overwrites existing assay
se2 <- minmax_normalization(se)

apply(assay(se2), 2, range)
#>      SRR1039508 SRR1039509 SRR1039512 SRR1039513 SRR1039516 SRR1039517
#> [1,]          0          0          0          0          0          0
#> [2,]          1          1          1          1          1          1
#>      SRR1039520 SRR1039521
#> [1,]          0          0
#> [2,]          1          1


# If new new_assay_name, normalization stored in a new object
se2 <- minmax_normalization(se, new_assay_name = 'minmax_counts')

apply(assay(se2, 'minmax_counts'), 2, range)
#>      SRR1039508 SRR1039509 SRR1039512 SRR1039513 SRR1039516 SRR1039517
#> [1,]          0          0          0          0          0          0
#> [2,]          1          1          1          1          1          1
#>      SRR1039520 SRR1039521
#> [1,]          0          0
#> [2,]          1          1

# A specific assay can also be selected
new_matrix <-  matrix(data = sample(x = seq(1, 100000),
                                   size = nrow(se) * ncol(se),
                                   replace = TRUE),
                     nrow = nrow(se),
                     ncol = ncol(se))
rownames(new_matrix) <- rownames(se)
colnames(new_matrix) <- colnames(se)

## Creating a new assay called new counts
assay(se, 'new_counts') <- new_matrix

se2 <- minmax_normalization(se,
                           new_assay_name = 'minmax_counts_new',
                           assay_name = 'new_counts')

apply(assay(se2, 'minmax_counts_new'), 2, range)
#>      SRR1039508 SRR1039509 SRR1039512 SRR1039513 SRR1039516 SRR1039517
#> [1,]          0          0          0          0          0          0
#> [2,]          1          1          1          1          1          1
#>      SRR1039520 SRR1039521
#> [1,]          0          0
#> [2,]          1          1

## Using a different range
se2 <- minmax_normalization(se,
                           new_assay_name = 'minmax_counts_new',
                           assay_name = 'new_counts',
                           new_min = 10,
                           new_max = 20)

apply(assay(se2, 'minmax_counts_new'), 2, range)
#>      SRR1039508 SRR1039509 SRR1039512 SRR1039513 SRR1039516 SRR1039517
#> [1,]         10         10         10         10         10         10
#> [2,]         20         20         20         20         20         20
#>      SRR1039520 SRR1039521
#> [1,]         10         10
#> [2,]         20         20