Type: Package
Title: Integrated Discovery of Oncogenic Signatures
Version: 1.0.1
Date: 2024-03-09
Author: Syed Haider [aut, cre], Francesca Buffa [aut]
Maintainer: Syed Haider <Syed.Haider@icr.ac.uk>
Depends: R (≥ 3.6.0), VennDiagram (≥ 1.6.5)
Description: A method to integrate molecular profiles of cancer patients (gene copy number and mRNA abundance) to identify candidate gain of function alterations. These candidate alterations can be subsequently further tested to discover cancer driver alterations. Briefly, this method tests of genomic correlates of mRNA dysregulation and prioritise those where DNA gains/amplifications are associated with elevated mRNA expression of the same gene. For details see, Haider S et al. (2016) "Genomic alterations underlie a pan-cancer metabolic shift associated with tumour hypoxia", Genome Biology, https://pubmed.ncbi.nlm.nih.gov/27358048/.
License: GPL-2
LazyLoad: yes
RoxygenNote: 7.2.3
NeedsCompilation: no
Packaged: 2024-03-09 20:20:57 UTC; shaider
Repository: CRAN
Date/Publication: 2024-03-11 18:50:09 UTC

create.counts.table

Description

Summary function to collapse the counts of selected (e.g. correlated) features per cancer type into counts table

Usage

create.counts.table(corr.summary = NULL)

Arguments

corr.summary

A list object containing subtype specific selected (e.g. correlated) features. This is the list object returned by estimate.expression.cna.correlation

Value

A matrix of cancer type specific counts

Author(s)

Syed Haider

See Also

estimate.expression.cna.correlation

Examples


# load test data
x <- get.test.data(data.types = c("mRNA.T", "CNA"));

# temporary output directory
tmp.output.dir <- tempdir();

# go through each cancer type iteratively and perform mRNA-CNA correlation analysis
correlated.features <- list();
for (cancer.type in names(x$mRNA.T)) {

  # estimate mRNA and CNA correlation for each cancer/disease type
  correlated.features[[cancer.type]] <- estimate.expression.cna.correlation(
    exp.data = x$mRNA.T[[cancer.type]], 
    cna.data.log2 = x$CNA.log2[[cancer.type]], 
    corr.threshold = 0.3, 
    corr.direction = "two.sided", 
    subtypes.metadata = list(
      "subtype.samples.list" = list("All" = colnames(x$mRNA.T[[cancer.type]]))
      ), 
    feature.ids = rownames(x$mRNA.T[[cancer.type]]), 
    cancer.type = cancer.type, 
    data.dir = paste(tmp.output.dir, "/data/", cancer.type, sep = ""),
    graphs.dir = paste(tmp.output.dir, "/graphs/", cancer.type, sep = "")
    );
  }

# create counts table across cancer types
counts.table <- create.counts.table(corr.summary = correlated.features);

create.training.validation.split

Description

Utility function to create random partitions of a dataset into training and validation sets. If samples are < 200, 66:34; otherwise 50:50 partitions are generated between training and validation sets respectively

Usage

create.training.validation.split(
  exp.data = NULL, ann.data = NULL, seed.number = 51214
  )

Arguments

exp.data

Feature by sample mRNA abundance matrix

ann.data

Sample by clinical attribute matrix

seed.number

Random seed for sampling

Value

A list of four matrices expression and two associated clinical matrices (exp.T, ann.T, exp.V and ann.V). One set for training and one for validation

Author(s)

Syed Haider

Examples


# load test data
x <- get.test.data(data.types = c("mRNA.T", "ann"));

# create training and validation sets
partitioned.datasets <- create.training.validation.split(
  exp.data = x$mRNA.T$BLCA, 
  ann.data = x$ann$BLCA, 
  seed.number = 51214
  );

estimate.expression.cna.correlation

Description

Estimate subtype specific correlation between mRNA and CNA profiles

Usage

estimate.expression.cna.correlation(
  exp.data = NULL, 
  cna.data.log2 = NULL, 
  corr.threshold = 0.3, 
  corr.direction = "two.sided", 
  subtypes.metadata = NULL, 
  feature.ids = NULL, 
  cancer.type = NULL, 
  data.dir = NULL, 
  graphs.dir = NULL
  )

Arguments

exp.data

Feature by sample mRNA abundance matrix

cna.data.log2

Feature by sample CNA log ratio matrix

corr.threshold

Threshold for Spearman's Rho to consider a feature as candidate driver

corr.direction

Whether to include positively (greater), negatively (less) or both (two.sided) correlated features. Defaults to two.sided

subtypes.metadata

Subtypes metadata list of lists. Must contain at least one subtype specific samples using list subtype.samples.list. If no subtypes are present, specify list element "All" with all samples

feature.ids

Vector of features to be used to estimate correlation

cancer.type

Name of the cancer type or dataset

data.dir

Path to output directory where mRNA and CNA correlation statistics will be stored

graphs.dir

Path to graphs directory

Value

A list of lists containing correlated features per cancer subtype

Author(s)

Syed Haider

Examples


# load test data
x <- get.test.data(data.types = c("mRNA.T", "CNA"));

# temporary output directory
tmp.output.dir <- tempdir();

# estimate mRNA and CNA correlation
correlated.features <- estimate.expression.cna.correlation(
  exp.data = x$mRNA.T$BLCA, 
  cna.data.log2 = x$CNA.log2$BLCA, 
  corr.threshold = 0.3, 
  corr.direction = "two.sided", 
  subtypes.metadata = list(
    "subtype.samples.list" = list("All" = colnames(x$mRNA.T$BLCA))
    ), 
  feature.ids = rownames(x$mRNA.T$BLCA), 
  cancer.type = "BLCA", 
  data.dir = paste(tmp.output.dir, "/data/BLCA/", sep = ""),
  graphs.dir = paste(tmp.output.dir, "/graphs/BLCA/", sep = "")
  );

estimate.null.distribution.correlation

Description

Function to estimate probability of observing correlations as high as observed using a feature list of interest

Usage

estimate.null.distribution.correlation(
  exp.data = NULL, 
  cna.data.log2 = NULL, 
  corr.threshold = 0.3, 
  corr.direction = "two.sided", 
  subtypes.metadata = NULL, 
  feature.ids = NULL, 
  observed.correlated.features = NULL, 
  iterations = 50, 
  cancer.type = NULL, 
  data.dir = NULL
  )

Arguments

exp.data

Feature by sample mRNA abundance matrix

cna.data.log2

Feature by sample CNA log ratio matrix

corr.threshold

Threshold for Spearman's Rho to consider a feature as candidate driver

corr.direction

Whether to include positively (greater), negatively (less) or both (two.sided) correlated features. Defaults to two.sided

subtypes.metadata

Subtypes metadata list. Contains at least subtype specific samples

feature.ids

Vector of features to be used to estimate correlation

observed.correlated.features

List of features that were found to be correlated for subtypes of a given cancer type

iterations

Number of random permutations for estimating p value

cancer.type

Name of the cancer type or dataset

data.dir

Path to output directory where the randomisation results will be stored

Value

1 if successful

Author(s)

Syed Haider

See Also

estimate.expression.cna.correlation

Examples


# load test data
x <- get.test.data(data.types = c("mRNA.T", "CNA"));

# temporary output directory
tmp.output.dir <- tempdir();

# estimate mRNA and CNA correlation for each cancer/disease type
correlated.features <- estimate.expression.cna.correlation(
  exp.data = x$mRNA.T$BLCA, 
  cna.data.log2 = x$CNA.log2$BLCA, 
  corr.threshold = 0.3, 
  corr.direction = "two.sided", 
  subtypes.metadata = list(
    "subtype.samples.list" = list("All" = colnames(x$mRNA.T$BLCA))
    ), 
  feature.ids = rownames(x$mRNA.T$BLCA), 
  cancer.type = "BLCA", 
  data.dir = paste(tmp.output.dir, "/data/BLCA/", sep = ""),
  graphs.dir = paste(tmp.output.dir, "/graphs/BLCA/", sep = "")
  );

# estimate NULL distribution
estimate.null.distribution.correlation(
  exp.data = x$mRNA.T$BLCA,
  cna.data.log2 = x$CNA.log2$BLCA, 
  corr.threshold = 0.3, 
  corr.direction = "two.sided", 
  subtypes.metadata = list(
    "subtype.samples.list" = list("All" = colnames(x$mRNA.T$BLCA))
    ), 
  feature.ids = rownames(x$mRNA.T$BLCA), 
  observed.correlated.features = correlated.features$correlated.genes.subtypes, 
  iterations = 50, 
  cancer.type = "BLCA", 
  data.dir = paste(tmp.output.dir, "/data/BLCA/", sep = "")
  );

find.DE.features

Description

Funtion to identify differentially expressed/variable features between Tumour (T) and Normal (N) profiles

Usage

find.DE.features(
  exp.data.T = NULL, 
  exp.data.N = NULL, 
  feature.ids = NULL,
  test.name = "t.test"
  )

Arguments

exp.data.T

Feature by sample mRNA abundance matrix; tumour samples

exp.data.N

Feature by sample mRNA abundance matrix; normal/baseline samples

feature.ids

Vector of features to be used to estimate correlation

test.name

Specify the statistical test name (exactly as it appears in R). Supported tests are t.test, wilcox.test, var.test

Value

Feature by cancer type matrix of log2 fold change (T vs N) and adjusted P values. P values are estimated through test.name

Author(s)

Syed Haider

See Also

t.test, wilcox.test, var.test

Examples


# load test data
x <- get.test.data(data.types = c("mRNA.T", "mRNA.N"));

# list of features to be assessed for differential expression
feature.ids <- rownames(x$mRNA.T$BLCA);

DE.results <- find.DE.features(
  exp.data.T = x$mRNA.T, 
  exp.data.N = x$mRNA.N, 
  feature.ids = feature.ids,
  test.name = "t.test"
  );

get.program.defaults

Description

Get default datasets bundled with package for test runs

Usage

get.program.defaults()

Value

A list with program.data.dir containing path to example program directory and test.data.dir containing path to example datasets directory

Author(s)

Syed Haider

Examples

x <- get.program.defaults();

get.test.data

Description

Function to load test data

Usage

get.test.data(data.types = c("mRNA.T", "ann"))

Arguments

data.types

Datatypes to be read Valid datatypes are: mRNA.T, mRNA.N, CNA (includes: log2, calls and fractions), annotations

Value

List of lists containing datasets and respective molecular profiles as matrices

Author(s)

Syed Haider

Examples


x <- get.test.data(data.types = c("mRNA.T", "mRNA.N", "ann"));

get.top.features

Description

Prioritise top features satisfying the criteria specified by various parameters described below

Usage

get.top.features(
  DE.features = NULL, 
  cna.data.fractions = NULL, 
  mRNA.FC.up = 0, 
  mRNA.FC.down = 0, 
  mRNA.p = 0.05, 
  mRNA.top.n = NULL, 
  cna.fractions.gain = 0.2, 
  cna.fractions.loss = 0.2
  )

Arguments

DE.features

Matrix containing differentially expressed features with two columns: FC and P. P may contain adjusted P or raw

cna.data.fractions

Feature by cancer type matrix with CNA fractions

mRNA.FC.up

Log2 fold change threshold for selecting over-expressed features

mRNA.FC.down

Log2 fold change threshold for selecting under-expressed features

mRNA.p

P value threshold for selecting significantly differentially expressed features. Mutually exclusive to mRNA.top.n

mRNA.top.n

Top n differentially expressed features satisfying each of the fold change criteria. Mutually exclusive to mRNA.p

cna.fractions.gain

Threshold for selecting copy number gain/amplifications

cna.fractions.loss

Threshold for selecting copy number losses

Value

Vector of top features

Author(s)

Syed Haider

Examples


# load test data
x <- get.test.data(data.types = c("mRNA.T", "mRNA.N", "CNA"));

# list of features to be assessed for differential expression
feature.ids <- rownames(x$mRNA.T$BLCA);

# get differentially expressed features
DE.results <- find.DE.features(
  exp.data.T = x$mRNA.T, 
  exp.data.N = x$mRNA.N, 
  feature.ids = feature.ids,
  test.name = "t.test"
  );

# get top features
top.features <- get.top.features(
  DE.features = cbind("FC" = DE.results[, 1], "P" = DE.results[, 2]),
  cna.data.fractions = x$CNA.fractions$BLCA, 
  mRNA.FC.up = 0.25, 
  mRNA.FC.down = 0.25, 
  mRNA.p = 0.05, 
  mRNA.top.n = NULL, 
  cna.fractions.gain = 0.2, 
  cna.fractions.loss = 0.2
  );

load.datasets

Description

Function to load and systemise molecular datasets

Usage

load.datasets(
  data.dir = "./", 
  metadata = NULL, 
  data.types = c("mRNA.T", "ann")
  )

Arguments

data.dir

Path to base data directory or directory containing molecular profiles

metadata

Dataset by profile metadata matrix containing file names of the molecular profiles for different datasets

data.types

Datatypes to be read Valid datatypes are: mRNA.T, mRNA.N, CNA (includes: log2, calls and fractions), annotations

Value

List of lists containing datasets and respective molecular profiles as matrices

Author(s)

Syed Haider

Examples


# locate test data directory which comes with the package
data.dir <- paste(system.file("programdata/testdata/", package = "iDOS"), "/", sep = "");

# read meta data file
metadata <- read.table(
  file = paste(data.dir, "metadata.txt", sep = ""), 
  row.names = 1, 
  header = TRUE, 
  sep = "\t",
  stringsAsFactors = FALSE
  );

x <- load.datasets(
  data.dir = data.dir,
  metadata = metadata,
  data.types = c("mRNA.T", "mRNA.N", "ann")
  );