Negative design is a group of virtual screening methods that aims at weeding out compounds with undesired properties during the early stages of drug development. These methods are mainly designed to predict three important types of pharmacological properties: drug-likeness, frequent hitters, and toxicity. In order to achieve high screening efficiency, most negative design methods are physicochemical property-based and/or substructure-based rules or filters. Such methods have advantages of simplicity and good interpretability, but they also suffer from some defects such as inflexibility, discontinuity, and hard decision-making. In this review, the advances in negative design for the evaluations of drug-likeness, frequent hitters, and toxicity are outlined. In addition, the related Web servers and software packages developed recently for negative design are summarized. Finally, future research directions in this field are discussed.