When researchers find their promising lead series facing IP issues, they are faced with a tough decision. Do they change series—possibly to a candidate that doesn’t show the same level of activity against the target—or do they spend a great deal of time and money screening the totality of the corporate database for a new prospect? In this situation, scaffold hopping and virtual screening are effective and efficient ways to identify a new series to move the project forward.
Scaffold hopping for leads
Scaffold hopping, based on an understanding of the required pharmacophore, can provide novel chemistry for the backup series. This can be particularly effective for finding novel leads in a congested intellectual property (IP) space.
Scaffold hopping is a computational technique of replacing portions of molecules to create novel drug-like compounds with similar activity to the original.
The method involves choosing a portion of the starting molecule, often the central scaffold, for replacement. The scaffold-hopping software searches a database of hundreds of thousands of fragments for the best replacements.
The worth of the software depends on the algorithm used to evaluate the “best” matches. Different software tools use different approaches. Some use simple geometrical considerations and/or the presence of simple pharmacophore points while others use ligand similarity to rank the replacements. In all cases, the best matches are returned as possible candidates for synthesis.
Choosing the right compounds to progress is important as it can frequently take up to a week or more of lab time to synthesize a new compound. Results must be imaginative, yet realistic suggestions that enable users to advance the compounds that are most likely to succeed.
Get it right the first time
Deciding which compound to progress is a complex mixture of different factors, including: how likely is the compound to be active; how well does it meet the objectives of the experiment; how easy is it to make; does it have the right balance of physicochemical properties.
The primary question for users of a scaffold-hopping program is, “How likely are the suggested compounds to be active?” This is directly related to the scoring that is used by the program and should be as accurate as possible.
Using simple geometric constraints (i.e., does the new scaffold physically fit into the space created by the original) is too simple to give a hit list rich in active molecules. Equally, traditional pharmacophore points such as presence or absence of a hydrogen-bond acceptor cannot reproduce the myriad of electronic and steric effects of the original scaffold. On the other hand, judging new scaffolds based on a simple molecular similarity may produce actives, but these are likely to share the same problems as the starting series.
An alternative technique creates a highly detailed, information-rich pharmacophore based on the minute electronic and steric properties—the molecular interaction field—of the starting scaffold.1 Using geometric constraints as a filter to remove scaffolds that will not physically fit into the space of the original scaffold, molecules are then analyzed based on the field pharmacophores. The new molecules with the highest degree of similarity in this analysis are also the most likely to be active at the target.
This approach produces a hit list that is highly likely to be active. By choosing molecules that fit their synthetic scheme and physiochemical property profile, chemists waste less time and money trying to find a new lead series.
Finding novel bioisosteres
The process is best illustrated by an example. The metabotropic glutamate receptors (mGluR) have become popular and important targets in small-molecule drug discovery. In particular, there is increasing interest in mGluR5 antagonists in the treatment of anxiety, depression, pain, gastroesophageal acid reflux disease (GERD), Parkinson’s disease, epilepsy, and Fragile X syndrome (FXS). Several mGluR5 antagonists have entered clinical trials.
The allosteric binding site of mGluR5, located within the transmembrane region, is generally considered to be more likely to lead to effective medicines than other binding sites. One of the prototypical small-molecule mGluR5 allosteric antagonists is MPEP [2-methyl-6-(phenylethynyl) pyridine].
Using MPEP as a starting point, scientists using sparkV10 from Cresset searched for bioisosteric replacements that would reveal novel IP. sparkV10 uses field technology to find biologically equivalent replacements for key moieties in a molecule. Using the molecular field descriptors as the basis for the search makes it possible to find new structures in new chemical space while retaining similar biological activity.
Firstly, the 3D molecular field descriptors were calculated for MPEP (Figure 1). Then two different moieties were identified as candidates for fragment swapping.
Two separate experiments were performed on MPEP. Firstly, sparkV10 was used to search for replacements for the central alkyne. sparkV10 searches a database of up to 600,000 fragments for bioisosteres that exhibit similar shape and electronic properties when placed in the context of the final molecule. sparkV10 uses molecular interaction fields to represent the key binding interactions of a molecule giving a close approximation to the protein’s view of a potential ligand. The results are shown in the left column of Figure 2.
Secondly, a search was carried out for replacements for the pyrido-alkyne section of the molecule. The results are shown in the right column of Figure 2.
Both experiments resulted in unreported, novel structures, along with a significant number of previously reported actives. The position of the identified actives was irrespective of the 2D similarity of the final molecule to MPEP.
This example demonstrates the power of bioisosteric fragment swapping using molecular fields to generate new leads and potential new intellectual property, even in crowded therapeutic areas.
Virtual screening for fast results
An alternative approach to identifying a backup series is to screen existing compounds to find one that is likely to be active against the target of interest. This offers the opportunity to jump directly into completely new chemical series, rather than the stepwise approach of scaffold hopping.
While this might seem simple, most Big Pharma databases contain anywhere from hundreds of thousands of structures to several millions, translating to a cost of more than $1.5 million to run a high-throughput screen of the whole collection. Additionally, many smaller companies lack a collection of sufficient size to attempt this approach.
Virtual screening is an intelligent way to reduce the cost of dealing with such a large number of available molecules. Narrowing down a large database to a manageable number of compounds of interest can vastly reduce the cost of a screening campaign from millions of dollars down to thousands.
To run a virtual screen the molecular field of the biologically active compound is calculated to quantify its biological activity profile. This is used to define a template with the desired biological characteristics for a potential new series. This template—or pharmacophore—forms the basis for running a virtual screen to filter down the full compound database to a much smaller set for screening.
Screening on a manageable scale
Many computational companies offer an outsourcing service whereby they will run a virtual screen for on a one–off or contractual basis. Outsourcing a virtual screen to consultants can cost as little as $15,000. The virtual screen results in a focused library of, say, the 500 most interesting compounds, which will cost less than $15,000 to screen for most standard assays. Using computational chemistry in this way can result in a one-hundred fold reduction of costs.
Case study of 11βHSD-1
The enzyme 11β-hydroxysteroid dehydrogenase type 1(11βHSD-1) was originally thought to predominate as an oxidase, converting cortisol to cortisone. However, 11βHSD-1 has since been shown to act as a reductase in vivo for cortisone. This strongly suggests that the inhibition of 11βHSD-1 and the associated decrease of active cortisol could be important in the control of obesity, insulin resistant diabetes, and cognition.
A research group wanted to generate new ideas for lead structures in this area, but had only the natural ligand as a starting point. They asked Cresset to help them to find some possible new lead series. In the absence of specific X-ray data, the first step was to model the binding action of cortisol and cortisone. From this, it was possible to deduce the molecular field of the binding fragment.
This binding fragment was input into Cresset’s blazeV10 software. Using the shape and electrostatic character of the binding fragment, blazeV10 searched a database of millions of known structures to find compounds with a similar field pattern. Since ligand interactions are based on the shape and electrostatic character of molecules, compounds that share similar field patterns are likely to have similar biological activity, regardless of their chemical structure.
The results of the virtual screen were reviewed on the basis of best field and volume fit to its seed, as well as structural diversity, ease of chemical modification, and intuitive appeal. The best 500 hits were tested. The results of the wet screen showed that ten compounds were active at <10μM, one at 470nM and the best at 170nM. Three of the most active compounds are shown in Figure 4, illustrating the diverse chemotypes that virtual screening can identify.
The striking feature of both scaffold hopping and virtual screening using fields is the diversity of leads that they produce. Each method relies on the underlying algorithm using the molecular field patterns, rather than the structural similarity, to find molecules for potential new series. The results retain the essential activity at the key binding sites, but with novel chemical structures.
The results and the opportunities that these methods bring make them invaluable tools to chemists looking to identify new chemistry for backup series.
1. Cheeseright T. The identification of bioisosteres as drug development candidates. Innovations in Pharmaceutical Technology. 2009; 28:22-26.