Inferring latent heterogeneity using many feature variables supervised by survival outcome

Area of Science:

• Biostatistics
• Computational Biology
• Oncology

Background:

• Understanding cancer patient heterogeneity is crucial for effective precision medicine.
• Identifying latent patient subgroups can guide targeted therapies for high-risk individuals.

Purpose of the Study:

• To develop a statistical model for inferring latent heterogeneity in cancer patient survival.
• To incorporate variable selection for identifying key features characterizing these subgroups.

Main Methods:

• A mixture model approach is proposed to capture distinct patient survival patterns.
• Multinomial distribution models mixing probabilities for latent groups.
• Adaptive lasso is integrated for parsimonious variable selection.

Main Results:

• The proposed adaptive lasso estimator demonstrates oracle properties.
• Simulation studies validate the finite sample performance of the method.
• The model is successfully applied to two real-world cancer datasets.

Conclusions:

• The developed mixture model effectively identifies latent patient heterogeneity in cancer.
• Variable selection enhances the interpretability and applicability of the model for precision oncology.
• This approach aids in timely and precise targeting of high-risk cancer patients.