For the precision medicine, we propose a framework for sensitive-information-safe learning of personalized models of the patients, which are de-facto unique due to multiple, co-occuring diseases.