Computational identification of bioactive natural products by structure activity relationship

Natural products (NPs) have been widely used in traditional medicines and are a valuable source for new drug discovery. However, insufficient knowledge about their molecular mechanisms has limited the scope of their application and hindered the effort to design new drugs from their synergistic action strategies. Thus far, a systematic study of all NP ingredients in a traditional medicine recipe remains impractical. However encouraging results have begun to appear illustrating synergies between several principle active ingredients. In this work, we propose the use of structure activity relationship (SAR) to identify potential active ingredients in natural products, with the aim to facilitate experimental and computational characterizations of their therapeutic mechanisms and synergies. We call this approach the bioactive natural compound-likeness (BNC-likeness) approach, drawing a parallel to the concept of drug-likeness. In cross-validations and independent example tests, our approach displayed 90-92% sensitivity and 85-90% specificity, suggesting its practical usefulness. We also showed that BNC-like compounds were not just drug-like NP ingredients. BNC-like compounds and drug-like chemicals may share different structural characteristics. Therefore, BNC-likeness is a helpful novel conception inviting dedicated research.