Rob Tibshirani propose to use lasso with Cox regression for variable selection in his 1997 paper "The lasso method for variable selection in the Cox model" published in Statistics In Medicine 16:385. Does anyone know of any R package/function or syntax in R that does lasso with a Cox model?
Here are two suggestions. First, you can take a look at the glmnet package, from Friedman, Hastie and Tibshirani, but see their JSS 2010 (33) paper, Regularization Paths for Generalized Linear Models via Coordinate Descent.
Second, although I've never used this kind of penalized model, I know that the penalized package implements L1/L2 penalties on GLM and the Cox model. What I found interesting in this package (this was with ordinary regression) was that you can include a set of unpenalized variables in the model.
The associated publication is now:
Goeman J.J. (2010). L-1 Penalized Estimation in the Cox Proportional Hazards Model. Biometrical Journal 52 (1) 70-84.
Many years after the question was posed, of course, but it seems that there is a Coxnet R package (since 2015) https://cran.r-project.org/web/packages/Coxnet/Coxnet.pdf, which i plan to try out for Penalized Cox Model for proteomics data.