Multi-targeted single molecules have many advantages over multiple agents such as “drug cocktails”, linked drugs and others. In order to design in silico multi-targeted agents there is a need to provide models based on ligands, or based on protein targets’ structures, or both ¹. There are many possible combinations of requirements for targets and anti targets concerning agonists and antagonists of CB1 and CB2 in the CNS and in the periphery.

Peripheral treatment with a CB1 antagonist reduced food intake and body weight ². Selective CB2 activation may serve therapeutic manipulation of untoward immune responses including those associated with a variety of neuropathies that exhibit a hyper-inflammatory component ³ (devoid of psychoactive effects of CB1 activation).

Combination of targeting specific receptors as well as anti-targets, could be achieved using our in-house Iterative Stochastic Elimination (ISE) Algorithm⁴ for constructing activity models of the cannabinoid receptors. ISE models are based on the idea of classification but provide an ensemble of filters of molecular properties that enable the quantification of each molecule’s chances due to the differences in classification ability between filters. These models will form the basis for subsequent virtual screenings of huge libraries of commercially available molecules to filter candidates that will be purchased and tested⁵. Thus, ISE models serve to discover new multitargeting candidates.

 * Molecular Modeling and Drug Discovery Lab, Institute for Drug Research, The Hebrew University of Jerusalem, Israel

1. Boran, A. D. W. & Iyengar, R. Systems pharmacology. Mt. Sinai J. Med. 77, 333–44
2. Nogueiras, R. et al. Peripheral, but not central, CB1 antagonism provides food intake-independent metabolic benefits in diet-induced obese rats. Diabetes 57, 2977–91 (2008).
3. Cabral, G. A. & Griffin-Thomas, L. Emerging Role of the CB2 Cannabinoid Receptor in Immune Regulation and Therapeutic Prospects. Expert Rev. Mol. Med. 11, e3
4. Stern N and Goldblum A. Iterative Stochastic Elimination for Solving Complex Combinatorial Problems in Drug Discovery Isr. J. Chem. 54, 1338-67 (2014)
5. Zatsepin M et al. Computational Discovery and Ecperimnetal Confirmation of TLR9 Receptor Antagonist Leads, http://pubs.acs.org/doi/abs/10.1021/acs.jcim.6b00070

Our laboratory developed a unique chemoinformatics classification method for modeling and discovery of novel drug candidates. We transform molecular 2D structures into a large set of physico-chemical properties, and use biological acitivy data from specific websites in order to construct “learning sets” that contain known active molecules which are “diluted” with a large number of inactive molecules. We focus mainly on “ligand based” modeling, which is not dependent on the structure of protein targets and are particularly useful if such target structures are absent¹.

Iterative Stochastic Elimination (ISE)² algorithm was devised to solve combinatorial “explosive” problems with frequently more than 10¹⁰⁰ possible combinations. ISE identifies those variable values that contribute consistently to the worst results, but do not contribute similarly to the best solutions, and eliminates them thus producing a smaller set of variable values and a smaller number of combinatorial options, allowing one to proceed to the discovery of a set of optimal states of the system – an ensemble of filters of molecular properties (calculated by MOE software³, each filter composed of a few (4-5) molecular physicochemical properties and their ranges.  The final set of "best filters" solutions is considered to be a “model”.  For both receptors (CB1 and CB2), We construct models of agonism and of antagonism, each one separately, for both CB1 and CB2, based on resources of data such as ChEMBL⁴. These are used subsequently to score any molecule for its ability to have any type of interaction with CB1 and CB2, and thus to discover novel cannabinoids with specific activities.

 * Molecular Modeling and Drug Discovery Lab, Institute for Drug Research, The Hebrew University of Jerusalem, Israel

1. Li, H. et al. Machine learning approaches for predicting compounds that interact with therapeutic and ADMET related proteins. J. Pharm. Sci. 96, 2838–60 (2007).
2. Stern, N. & Goldblum, A. Iterative Stochastic Elimination for Solving Complex Combinatorial Problems in Drug Discovery. Isr. J. Chem. 54, 1338–1357 (2014).
3. Molecular Operating Environment (MOE), 2012.10; Chemical Computing Group Inc., 1010 Sherbooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2012.
4. https://www.ebi.ac.uk/chembl/