Skip to contents

martini

martini provides a convenient framework to collect data from multiple clinical trial data sets following the ADaM standard and create a machine-learning-ready data set.

It is part of a set of packages that form a pipeline for machine learning and artificial intelligence applications in clinical research that offers a streamlined and standardized way to assess relationships between clinical study data and a specified outcome while aligning with validation and documentation standards expected for software used in clinical trials.

While designed to interface smoothly with internal downstream modeling and reporting tools, the outputs from martini can of course also be used with any ML framework of your choice.

The core of martini revolves around three key steps, each represent by one main function:

  • spec() – Specify the relevant steps to collect and combine the relevant data
  • build() – Build one raw analysis dataset with the relevant clinical features
  • prepare_ml() – Prepare clean, analysis-ready data object for modeling

Installation

To install the latest version of martini from GitHub:

remotes::install_git(
  "https://github.com/Bayer-Group/bmdi-mlai-martini",
  dependencies    = TRUE,
  build_vignettes = TRUE
)

Documentation

Please refer to the package vignettes for an exemplary analysis either by running the following code or by opening the
pkgdown site.

vignette(package = "martini")