Attempting to answer to this question, we developed a novel method to convert the pharmacologically-relevant information of natural products into numbers (molecular descriptors), which can be used for virtual screening, drug design and artificial intelligence applications. 

For millennia, nature has provided humans with medicaments in the form of plant extracts, ointments, potions and oils. The underlying active ingredients are constituted by natural products, which have evolved over millions of years as preferred molecular frameworks for biological interactions. Natural product scaffolds are privileged molecular cores for the successful design of novel drugs and represent an unparalleled source of structural information, with their complementary structures compared to man-made, synthetic chemicals [1-3]. However, natural products often possess intricate molecular structures, posing synthetic difficulties. So, how can we harness the natural chemical diversity for the discovery of compounds with an increased synthetical accessibility and greater drug-likeness? 

To answer to this question, chemoinformatics, which utilizes computer-assisted techniques to solve chemical challenges, can come into the chemists’ aid. In particular, the so-called molecular descriptors represent a solution to translate chemical information into useful numbers, thereby enabling virtual screening, computer-assisted drug design and machine-learning applications. 

By melding recent statistical advances into pharmacologically-relevant chemical information, e.g., atomic partial charges and 3-dimensional conformation, we have developed the Weighted Holistic Atom Localization and Entity Shape (WHALES) description. This computational method  encodes the patterns of shape, geometry and partial charges of natural products as numerical “fingerprints”. Such a molecular description can be used to identify synthetic compounds sharing the same patterns as the natural product templates, thereby potentially mimicking the same biological activity.

As the pioneering application of the novel method, we apply it to the discovery of synthetic bioactives utilizing four phytocannabinoids as templates. By screening a virtual library of more than 3 million compounds, we identified seven novel synthetic compounds that modulate cannabinoid 1 and cannabinoid 2 receptors in vitro. These innovative modulators are structurally less complex than their respective natural product templates, possess novel scaffolds and populate druglike regions of the chemical space. 

The novel computational approach, which is provided as an open-access toolkit (available here), bears promise as a chemoinformatic strategy for grasping the structural complexity of natural products, thereby enabling scaffold-hopping to synthetic mimetics. Additional information can be found in the original paper (DOI: 10.1038/s42004-018-0043-x).

[1] Lee, M. L. & Schneider, G. Scaffold architecture and pharmacophoric properties of natural products and trade drugs: application in the design of natural product-based combinatorial libraries. J. Comb. Chem. 3, 284–289 (2001).

[2] Henkel, T., Brunne, R. M., Müller, H. & Reichel, F. Statistical investigation into the structural complementarity of natural products and synthetic compounds. Angew. Chem. Int. Ed. 38, 643–647 (1999).

[3] Rodrigues, T., Reker, D., Schneider, P. & Schneider, G. Counting on natural products for drug design. Nature Chem. 8, 531-542 (2016).