combine.integrative.features {iSubGen}

Combine iSubGen integrative features

Description

Combine a independent reduced features matrix (ex. from autoencoders) and pairwise integrative similarity matrices into one integrative feature matrix.

Usage

combine.integrative.features(irf.matrix, cis.matrix,
irf.rescale.recenter = NA, cis.rescale.recenter = NA, 
irf.rescale.denominator = NA, cis.rescale.denominator = NA,
irf.weights = rep(1, ncol(irf.matrix)), 
cis.weights = rep(1, ncol(cis.matrix)))

Arguments

irf.matrix

matrix of independent reduced features with patients as rows and features as columns

cis.matrix

matrix of consensus integrative similarity or integrative similarity features with patients as rows and features as columns

irf.rescale.recenter

either NA, "mean", a single number or a vector of numbers of length equal to the number of columns of irf

cis.rescale.recenter

either NA, "mean", a single number or a vector of numbers of length equal to the number of columns of cis

irf.rescale.denominator

either NA, "sd", a single number or a vector of numbers of length equal to the number of columns of irf

cis.rescale.denominator

either NA, "sd", a single number or a vector of numbers of length equal to the number of columns of cis

irf.weights

single number or vector of numbers of length equal to the number of columns of irf

cis.weights

single number or vector of numbers of length equal to the number of columns of cis

Details

The recenter values determine the how column centering is performed. If NA, no recentering is done. If the values equal "mean", then the mean of each column will be used. Otherwise, the numeric values specified will be used. The denominator values determine how column scaling is performed. If NA, no recentering is done. If the denominator values equal "sd", then the standard deviation of each column will be used. Otherwise, the numeric values specified will be used. The values used are returned by the function along with the compressed feature matrix to be recorded for reproducibility purposes.

Value

integrative.feature.matrix

a matrix of compressed features with patients as rows and features as columns

irf.rescale.recenter

a numeric vector with length equal to the number of columns of irf

cis.rescale.recenter

a numeric vector with length equal to the number of columns of cis

irf.rescale.denominator

a numeric vector with length equal to the number of columns of irf

cis.rescale.denominator

a numeric vector with length equal to the number of columns of cis

irf.weights

a numeric vector with length equal to the number of columns of irf

cis.weights

a numeric vector with length equal to the number of columns of cis

Author(s)

Natalie Fox

Examples

# Create matrices for combining
irf.matrix <- matrix(runif(25*4), ncol = 4);
rownames(irf.matrix) <- c(paste0('EP00',1:9), paste0('EP0',10:25));
cis.matrix <- matrix(runif(25*6), ncol=6);
rownames(cis.matrix) <- c(paste0('EP00',1:9), paste0('EP0',10:25));

# Example 1: Join the matrices without any weighting adjustments
isubgen.feature.matrix <- combine.integrative.features(
  irf.matrix,
  cis.matrix
  )$integrative.feature.matrix;

# Example 2: Combine matrices after scaling each column by subtracting the mean
# and dividing by the standard devation of the column
isubgen.feature.matrix.rescaled.result <- combine.integrative.features(
  irf.matrix,
  cis.matrix,
  irf.rescale.recenter = 'mean',
  cis.rescale.recenter = 'mean',
  irf.rescale.denominator = 'sd',
  cis.rescale.denominator = 'sd'
  );
isubgen.feature.matrix.2 <- isubgen.feature.matrix.rescaled.result$integrative.feature.matrix;

# Example 3: Combine matrices 
isubgen.feature.matrix.reweighted.result <- combine.integrative.features(
  irf.matrix,
  cis.matrix,
  irf.weights = 1/4,
  cis.weights = 1/6
  );
isubgen.feature.matrix.3 <- isubgen.feature.matrix.reweighted.result$integrative.feature.matrix;