fit.interaction.model {SIMMS}

Cox model two features separately and together

Description

Using a meta-analysis dataset take two features and Cox model them separately and together and extract HRs and p-values.

Usage

fit.interaction.model(
  feature1,
  feature2,
  expression.data,
  survival.data,
  data.type.ordinal = FALSE,
  centre.data = "median"
)

Arguments

feature1

String indicate what feature (gene/probe/etc.) should be extracted for analysis

feature2

String indicate what feature (gene/probe/etc.) should be extracted for analysis

expression.data

A list where each component is an expression matrix (patients = columns, features = rows) for a different dataset

survival.data

A list where each component is an object of class Surv

data.type.ordinal

Logical indicating whether to treat this datatype as ordinal. Defaults to FALSE

centre.data

A character string specifying the centre value to be used for scaling data. Valid values are: 'median', 'mean', or a user defined numeric threshold e.g. '0.3' when modelling methylation beta values. This value is used for both scaling as well as for dichotomising data for estimating univariate betas from Cox model. Defaults to 'median'

Details

The interaction model compares cases where feature1 and feature2 concord (both high or both low) to those where they do not. That is, the model is y = x1 + x2 + (x1 == x2) and not the typical y = x1 + x2 + x1:x2

Value

Returns a vector of six elements containing (HR,P) pairs for feature1, feature2, and the interaction

Author(s)

Syed Haider & Paul C. Boutros

Examples

data.dir <- get.program.defaults()[["test.data.dir"]];
data.types <- c("mRNA");
x1 <- load.cancer.datasets(
  datasets.to.load = c('Breastdata1'),
  data.types = data.types,
  data.directory = data.dir
  );
x2 <- fit.interaction.model(
  feature1 = "1000_at", 
  feature2 = "2549_at",
  expression.data = x1$all.data[[data.types[1]]],
  survival.data = x1$all.survobj
  );