Regularized Latent Class Model for Joint Analysis of High-Dimensional Longitudinal Biomarkers and a Time-to-Event...

Area of Science:

• Biostatistics
• Genomics
• Computational Biology

Background:

• Joint analysis of longitudinal biomarkers and time-to-event data is crucial for understanding disease progression.
• Existing methods often struggle with high-dimensional biomarker data.

Purpose of the Study:

• To propose a novel joint latent class model for high-dimensional longitudinal biomarkers and survival outcomes.
• To improve prediction accuracy and subject classification in complex biomedical data.

Main Methods:

• Developed a joint latent class model with three components: class membership, survival, and longitudinal submodels.
• Employed a penalized likelihood approach to identify covariate effects on biomarkers and class membership.
• Utilized bootstrap for model inference.

Main Results:

• The proposed method demonstrated superior performance in prediction and class assignment compared to existing joint modeling techniques.
• Identified four distinct latent classes in idiopathic pulmonary fibrosis patients, with one class exhibiting significantly higher mortality risk.
• Revealed unique longitudinal gene expression trajectories for each latent class, offering new biological insights.

Conclusions:

• The novel joint latent class model effectively handles high-dimensional biomarker data for joint analysis with survival outcomes.
• The model provides accurate predictions and robust subject classification, aiding in personalized medicine.
• Application to idiopathic pulmonary fibrosis revealed distinct patient subgroups and identified novel gene expression patterns associated with survival.