Fitting high-dimensional mixture cure models using the hdcuremodelsR package

Area of Science:

• Biostatistics
• Computational Biology
• Genomics

Background:

• Time-to-event outcomes are crucial in biomedical research, especially with long-term survivors or uncured subjects.
• Mixture cure models (MCMs) are essential for such datasets.
• Identifying molecular features linked to time-to-event outcomes is vital for pathway elucidation and therapeutic targeting.

Purpose of the Study:

• Introduce the hdcuremodels R package for high-dimensional mixture cure modeling.
• Enable modeling of right-censored time-to-event data with a cured fraction and numerous predictors.
• Facilitate the identification of molecular features associated with survival outcomes.

Main Methods:

• Implemented expectation-maximization and generalized monotone incremental forward stagewise algorithms for model fitting.
• Developed cross-validation functions, with and without false discovery rate control.
• Included flexible modeling functions without predictor requirements for incidence and latency components.

Main Results:

• Demonstrated the fitting of a high-dimensional penalized mixture cure model.
• Applied the model to an acute myeloid leukemia dataset.
• Achieved strong predictive performance on an independent test set.

Conclusions:

• The hdcuremodels package effectively fits penalized mixture cure models.
• It accommodates datasets where the number of predictors exceeds the sample size.
• Provides tools for analyzing complex time-to-event data in high-dimensional biological studies.