Predicting drug-disease associations by using similarity constrained matrix factorization