I am running a glmnet model in R using the caret package and doing repeated nested cross-validation with the nestedcv package. I need to include a “custom” interaction of a single one of my categorical features (which has values 0/1) with all other features (some of which are numerical, others are categorical and coded as 0/1).