We designed a Regularisation-path algorithm that uses patterns of input variables (itemsets) as covariates to fit a Cox Proportional Hazards regression model. Full description of method and algorithm published in Oxford Journal's Bioinformatics (based on work by Park & Hastie (2006): L1 Regularization Path Algorithm for Generalized Linear Models).
The entire source code (in R) for this project, can be found on a public GitHub repository.
D. duVerle, I. Takeuchi, Y. Murakami-Tonami, K. Kadomatsu and K. Tsuda. Discovering Combinatorial Interactions in Survival Data. Bioinformatics, 29(23):3053-3059, 2013.