A Network-Constrain Weibull AFT Model for Biomarkers Discovery.

Area of Science:

• Biostatistics
• Computational Biology
• Statistical Modeling

Background:

• Survival analysis is crucial for modeling time-to-event data in various scientific fields.
• Existing methods may not fully leverage complex predictor relationships or perform robust variable selection.
• Accelerated Failure Time (AFT) models offer an alternative to proportional hazards models.

Purpose of the Study:

• To propose AFTNet, a novel network-constrained survival analysis method.
• To enhance variable selection and estimation within the Weibull AFT framework.
• To incorporate predictor correlations for improved model performance.

Main Methods:

• Developed AFTNet, a network-constraint survival analysis method.
• Utilized a penalized likelihood approach for variable selection and estimation.
• Employed a double penalty to address structured sparse regression, promoting sparsity and grouping effects.
• Established theoretical consistency of the AFTNet estimator.
• Implemented an efficient iterative algorithm based on proximal gradient descent.

Main Results:

• AFTNet effectively performs variable selection and estimation in survival analysis.
• The double penalty successfully incorporates predictor correlation patterns.
• Theoretical consistency of the AFTNet estimator was demonstrated.
• The computational algorithm proved efficient on synthetic and real data.

Conclusions:

• AFTNet provides a robust and efficient method for network-constrained survival analysis.
• The method offers advantages in variable selection and handling correlated predictors.
• AFTNet shows promising performance on both simulated and real-world datasets.