calculate.cis.matrix {iSubGen}

Calculate consensus integrative correlation matrix

Description

Calculate consensus pairwise correlations between patient distances

Usage

calculate.cis.matrix(data.types, data.matrices, dist.metrics,
correlation.method = "spearman", filter.to.common.patients = FALSE, 
patients.to.return = NULL, patients.for.correlations = NULL, 
patient.proportion = 0.8, feature.proportion = 1, num.iterations = 10, 
print.intermediary.similarity.matrices.to.file = TRUE, print.dir = '.',
patient.proportion.seeds = seq(1,num.iterations), 
feature.proportion.seeds = seq(1,num.iterations))

Arguments

data.types

vector of the IDs for the different data types that are the names of the lists for the data.matrices and dist.metrics

data.matrices

list of the matrices with features (rows) by patients (columns)

dist.metrics

list of the distance metrics for comparing patient profiles. ex. euclidean. Options are from philentropy::distance

correlation.method

specifies the type of correlation for similarity comparison. Options are pearson, spearman or kendall.

filter.to.common.patients

logical, where TRUE indicates to filter out patients that don't have all data types

patients.to.return

vector of patients to calculate CIS for. For example, this is the testing cohort patients when calculating CIS for the testing cohort using the training cohort patients. If NULL all patients/columns will be used.

patients.for.correlations

vector of patients to use to calculate the similarities. For example, this would be the training cohort patients when calculating CIS for the testing cohort. If NULL all patients/columns will be used.

patient.proportion

proportion of patients.for.correlations to sample for each iteration (sampled without replacement).

feature.proportion

proportion of the features to sample for each iteration (sampled without replacement).

num.iterations

number of iterations to take the median from

print.intermediary.similarity.matrices.to.file

logical, where TRUE indicates that created intermediary integrative similarity matrix from each iteration should be printed to file

print.dir

directory for where to print the intermediary similarity matrices to file

patient.proportion.seeds

vector of scalars of the length num.iterations specifying the seeds used for random sampling for selecting the patient subsets at each iteration

feature.proportion.seeds

vector of scalars of the length num.iterations specifying the seeds used for random sampling for selecting the feature subsets at each iteration

Value

CIS matrix where rows are patients and columns are pairs of data types

Author(s)

Natalie Fox

Examples

# Load molecular profiles for three data types from example files saved 
# in the package as <data type>_profiles.txt
example.molecular.data.dir <- paste0(path.package('iSubGen'),'/exdata/');
molecular.data <- list();
for(i in c('cna','snv','methy')) {
  molecular.data[[i]] <- load.molecular.aberration.data(
    paste0(example.molecular.data.dir,i,'_profiles.txt'),
    patients = c(paste0('EP00',1:9), paste0('EP0',10:30))
    );
  }

# Example 1: calculate the consensus integrative similarity (CIS) matrix
corr.matrix <- calculate.cis.matrix(
  data.types = names(molecular.data),
  data.matrices = molecular.data,
  dist.metrics = list(
    cna = 'euclidean',
    snv = 'euclidean',
    methy = 'euclidean'
    ),
  print.intermediary.similarity.matrices.to.file = FALSE
  );

# Example 2: calculate the CIS matrix for patients EP001 through EP009 in relation 
# to patients EP010 through EP030 meaning the profile of EP001 is correlated to 
# the profiles of EP010 through EP030 so when assessing new patients, they can be 
# compared to the training profiles
corr.matrix2 <- calculate.cis.matrix(
  data.types = names(molecular.data),
  data.matrices = molecular.data,
  dist.metrics = list(
    cna = 'euclidean',
    snv = 'euclidean',
    methy = 'euclidean'
    ),
  patients.to.return = paste0('EP00',1:9),
  patients.for.correlations = paste0('EP0',10:30),
  print.intermediary.similarity.matrices.to.file = FALSE
  );

# Example 3: Adjusting the proportion of the features that will be used to correlate 
# the patient profiles
corr.matrix3 <- calculate.cis.matrix(
  data.types = names(molecular.data),
  data.matrices = molecular.data,
  dist.metrics = list(
    cna = 'euclidean',
    snv = 'euclidean',
    methy = 'euclidean'
    ),
  patients.to.return = paste0('EP00',1:9),
  patients.for.correlations = paste0('EP0',10:30),
  feature.proportion = 0.6,
  print.intermediary.similarity.matrices.to.file = FALSE
  );