OncoCast R package

Please find all the package’s resources on it’s github repository.

An Ensemble Learning Approach for Outcome Prediction in Precision Oncology Setting

Ensemble learner framework for survival outcome prediction and stratification for high dimension data. Originally developed for cancer genomics with delayed entry in the risk set in mind, and thus can adjust for left-truncation. OncoCast enables users to easily perform one or multiple machine learning survival analyses at once and explore and visualize the resulting output.

Installing OncoCast

OncoCast has some dependencies that will be installed if they are not found in your library. When installing from github a prompt in the console may ask if you want to install the binaries for curl v4.0 instead of 3.3. There is no need to update it for the package to work properly.

install.packages("remotes")
remotes::install_github("AxelitoMartin/OncoCast")

If you wish to use the development version of the package please use:

remotes::install_github("AxelitoMartin/OncoCast", ref = "development")

Using OncoCast

We recommend users to walk through the companion website to this package before their first use of the method. They will be guided through:

• Generating data in a ready for analysis format where the users will be guided on how to format their data to use OncoCast.
• Performing an ensemble learning run through a guide of the different machine learning algorithms available and how to create an ensemble model.
• Generating comprehensible results with a tutorial on how to explore the predicted risk score of patients in the dataset, the prognostic power of the ensemble model, the importance of each feature of interest and how to optimize creation of risk groups in clinically meaningful subsets.
• Using web-based interactive applications to simplify the exploration of results and sharing them collaborators.

OncoCast Online

There exist a version of OncoCast completely web-based requiring no coding skills and minimal inputs to create and explore an ensemble model. It can be found through this online RShiny application. The user will only be asked to input the dataset they wish to study and the method they want to use to create the emsemble model.