In broadinstitute/cdsr_models: R toolkit for modeling

knitr::opts_chunk$set(warning = FALSE, message = FALSE, fig.width = 7, fig.height = 7, cache = T)

library(tidyverse)
library(useful)
library(taigr)
library(cdsrbiomarker)

cdsrmodels contains modeling function created by the cancer data science team.

Install

library(devtools)
devtools::install_github("broadinstitute/cdsr_models")

The package can then be loaded by calling

library(cdsrmodels)

Modeling functions

discrete_test

Compares binary features, such as lineage and mutation, running a t-test on the difference in mean response between cell lines with the feature and without it. Run on response vector y and feature matrix X

cdsrmodels::discrete_test(X, y)

lin_associations

Compares continuous features, such as gene expression, calculating correlations between response and each feature. Run on feature matrix A, response vector y, and an optional matrix of confounders W. Other parameters can also be tuned and are explained in the function documentation.

cdsrmodels::lin_associations(A, y, W=NULL)

random_forest

Fits a random forest to a feature matrix X and a response vector y returning estimates of variable importance for each feature, as well as model level statistics such as R-squared. Other parameters can also be tuned and are explained in the function documentation.