NetSurvProx

Introduces a novel network-constrained survival analysis framework for variable selection and parameter estimation in penalized survival models with convex penalties. The package extends two classical survival models, the Cox Proportional Hazards (PH) model and the Accelerated Failure Time (AFT) model, by incorporating prior biological knowledge from curated interaction networks (e.g., KEGG) into a double-penalty framework. The first penalty enforces variable selection through a LASSO penalty, while the second preserves gene-gene correlations by incorporating Laplacian-based constraints, ensuring that biologically relevant network structures are maintained. Using censored survival data, the method enables the identification of predictive biomarkers and pathways with potential relevance for target therapies. Model estimation is performed via proximal optimization algorithms combined with cross-validation for reliable tuning. To enhance interpretability, dedicated utility functions are implemented to consolidate results, yielding biologically coherent insights that can support personalized medicine and contribute to improved patient outcomes.

Examples that run

20%

Documented parameters

99%

Return-value docs

100%

References docs

0%

Nothing depends on this yet.

1 release. R releases are shown for context.

First published

2026-06-09

Total releases

1 / 1 yrs

License

GPL (>= 3) OSI

Minimum R

≥ 4.3

Bundled data

367 KB / 1 file

Download size

not tracked yet

Installed size

not tracked yet

With dependencies

not tracked yet