Natural products and their molecular frameworks have a long tradition as valuable starting points for medicinal chemistry and drug discovery. Recently, there has been a revitalization of interest in the inclusion of these chemotypes in compound collections for screening and achieving selective target modulation. Here we discuss natural-product-inspired drug discovery with a focus on recent advances in the design of synthetically tractable small molecules that mimic nature's chemistry. We highlight the potential of innovative computational tools in processing structurally complex natural products to predict their macromolecular targets and attempt to forecast the role that natural-product-derived fragments and fragment-like natural products will play in next-generation drug discovery.

No abstract

De novo molecular design and in silico prediction of polypharmacological profiles are emerging research topics that will profoundly affect the future of drug discovery and chemical biology. The goal is to identify the macromolecular targets of new chemical agents. Although several computational tools for predicting such targets are publicly available, none of these methods was explicitly designed to predict target engagement by de novo-designed molecules. Here we present the development and practical application of a unique technique, self-organizing map-based prediction of drug equivalence relationships (SPiDER), that merges the concepts of self-organizing maps, consensus scoring, and statistical analysis to successfully identify targets for both known drugs and computer-generated molecular scaffolds. We discovered a potential off-target liability of fenofibraterelated compounds, and in a comprehensive prospective application, we identified a multitarget-modulating profile of de novo designed molecules. These results demonstrate that SPiDER may be used to identify innovative compounds in chemical biology and in the early stages of drug discovery, and help investigate the potential side effects of drugs and their repurposing options.drug design | target prediction | polypharmacology | machine learning | chemical similarity C omputer-assisted de novo molecular design has evolved as a popular source of ideas to combat the perceived lack of new chemical entities (NCEs) in chemical biology and drug discovery (1). We demonstrate that automated de novo design delivers readily synthesizable NCEs with desirable activity profiles. Although receptor-based design operates on a model of a macromolecular binding site, ligand-based methods are either explicitly or implicitly based on the chemical similarity principle (2) without requiring a receptor model (3). Instead, the latter typically uses some measure of chemical or pharmacophore feature similarity to a reference ligand as a fitness function, which aims to generate NCEs as template mimetics via scaffold hopping (4-8). We report the development, implementation, and successful prospective application of an innovative computational technique for the target profiling of de novo-designed molecules. The approach combines the concepts of self-organizing maps (SOMs) (9) for macromolecular target prediction (10), consensus scoring (11), and a statistical evaluation that provides confidence estimates for the predictions.Predicting polypharmacological activities is a topic relevant to chemical biology and drug discovery, not only to take advantage of inherent drug promiscuity but also to decrease lead compound attrition caused by unfavorable off-target modulation

No abstract

Aim: Computational chemogenomics models the compound–protein interaction space, typically for drug discovery, where existing methods predominantly either incorporate increasing numbers of bioactivity samples or focus on specific subfamilies of proteins and ligands. As an alternative to modeling entire large datasets at once, active learning adaptively incorporates a minimum of informative examples for modeling, yielding compact but high quality models. Results/methodology: We assessed active learning for protein/target family-wide chemogenomic modeling by replicate experiment. Results demonstrate that small yet highly predictive models can be extracted from only 10–25% of large bioactivity datasets, irrespective of molecule descriptors used. Conclusion: Chemogenomic active learning identifies small subsets of ligand–target interactions in a large screening database that lead to knowledge discovery and highly predictive models.