The impact of high-throughput catalyst screening in drug discovery

The recent trend of chemical catalysis centers and industry-academia partnerships has led to an influx of new reactions with direct applications for medicinal chemists1.  High-throughput experimentation (HTE) approaches have significantly lowered sample and time requirements necessary to screen thousands of unique reaction combinations for the discovery of new catalysts and reactions.  Recent advances in photochemistry2, late-stage functionalization3 and generating scaffold diversity4 are intriguing new possibilities to access structural complexity, while classical favorites like Suzuki-Miyaura and Buchwald-Hartwig cross-coupling reactions have seen continued advances in catalyst design and reaction protocols5. Certainly, many medicinal chemists would like to incorporate new technology and chemistry into their programs; however, testing and implementing new techniques on early discovery timelines and budgets can be difficult.  With many options available to plan and execute the synthesis of a new series, the medicinal chemist remains tasked with producing testable amounts of unique material on short timelines.   How best can the medicinal chemist leverage new technology and chemistry to bring value to their program? How can small and mid-sized medicinal chemistry programs compete without millions in upfront investment to establish their own catalysis centers? To start to answer these questions, it is best to start at the beginning of the typical early discovery project. 

Implement New Chemical Technology

 

Modern chemistry has afforded screening libraries containing millions (billions!) of compounds which are tested for activity using high-throughput screening (HTS) techniques.    From this list of hits, the medicinal chemist has many options when setting a path forward for synthesis of a new series where diverse compounds are needed to fill out chemical space.  Molecular modeling and cheminformatics tools give the designer varied approaches to propose structures with a virtually unlimited array of predicted properties6.  It is a certainty that the designer can plan a list of targets faster than they can be synthesized. On paper (or a computer screen) all structures seem accessible and together serve to answer the biological hypothesis proposed.

A common total synthesis trope rephrased in many ways, “with enough time, energy and money any structure can be synthesized”. The field of natural product total synthesis has validated this assertion as some natural products, like observed for the successful synthesis of Maoecrystal V, required upwards of a thousand screening steps to find a suitable reaction for just one step of the scheme7.  With unlimited time, budget and robotics (or graduate students), one could test nearly every possible combination of known catalysts and reagents to find a working synthesis for every structure in development. This is, however, of no comfort to those at the forefront of early discovery where timelines and budgets are short with access to equipment often limited.

Each failed reaction is a missed opportunity

 

In early discovery, synthetic strategies which aspire to build structural complexity in as few reactions steps as possible are preferred.  Many strategies will invoke designing fragments of the molecule with connecting the structure in a late step with a cross-coupling reaction like a Suzuki or a Buchwald-Hartwig amination. Achieving the diversity proposed in the series design, will likely require many reactions with a diverse array of coupling partners, wildly varied reactivity and potentially incompatible functional groups, testing the effectiveness of these common coupling reactions.  As a first pass, a single catalyst and protocol may be selected to screen a diverse set of functionalized building blocks. Many of these reactions will undoubtedly work the first time and no additional reaction optimization will be necessary.  In early discovery, only a few mg of compound may be necessary for initial screening/testing and this may be accessible with minimal optimization.  However, an internal analysis of electronic notebook data for synthetic plans at Merck found that 55% of metal catalyzed C-N forming reactions in what were to be final steps of synthesis found no product at all8.  How to resolve this disconnect between the utility of the common cross-coupling reactions and the preponderance of failed reactions in real world scenarios? What implications do the missed reactions have on the success or failure of a project?

figure_1

Each failed reaction is a missed opportunity in a carefully planned series design. In a chemical space analysis performed by a research group at Merck, the majority of drug compounds based on cross-coupling chemistry fell far outside the chemical space covered by substrates in the initial methodology literature8. The structures most likely to be valuable from a medicinal chemistry perspective fall well outside the scope of successful reactions in the literature as presented and will likely require reaction optimization.  If too narrow a spectrum of catalysts is attempted or too little effort to resolve a problem reaction, you may miss a product which could be your next drug.  Extensive efforts are needed to improve the success rates for synthetic methodology on biologically relevant molecules to be useful in medical chemistry.

From Literature To The Chemist Bench

 

Literature methods are often reported using simple substrates demonstrating the high yield and utility of a new catalyst or reaction protocol. Comparing literature reports which rarely use the same substrate, scale, solvent or protocol makes direct comparison very difficult.  HTE tools enable analysis of reaction robustness for real world substrates on a standardized scale in broad depth in ways that are extremely difficult without automated tools.    A Merck group published a detailed analysis of cross-coupling chemistry using what they have termed Chemistry Informer Libraries8.  Using a pre-selected array of 24 boronate substrates, the group tested first generation palladium catalysts for the Suzuki coupling of a “simple” (aryl) bromide.  For 24 substrates, using the original Suzuki protocol and catalyst, Pd(Ph3P)4, 22 of 24 reactions gave higher than 20% with an average yield of 46%.  Switching to an aryl chloride with the same set of boronates, more variation was observed.  For the chloride, using Xphos G2 pre-catalyst, 20 of 24 reactions were greater than 20% for an average of 57% including 4 examples with greater than 98%. Interestingly, for the more evolved Xphos catalyst, 3 of the reactions report 0 % conversion. By changing temperature and other reaction parameters they show even more variation and selectivity.  It is easy to imagine the complexity that could be expected when expanding this analysis to additional Suzuki catalysts and more complex coupling partners.  However, by using a standard set of substrates this method of analysis allows direct comparison of reaction protocols.

 

figure_2

In the same report but for C-N bond formation, 8 catalyst and base combinations were screened with an informer library of 18 halides with a secondary amine.  The first condition reported by Buchwald and Hartwig using ((P(o-Tol)3)2PdCl2 was successful in only 2 of the 18 reactions at greater than 20%.  The most successful catalyst RuPhos G2 precat with Cs2CO3 gave only 6 reactions greater 20%.  For the 8 conditions tested, successful reactions ranged from 2-6 trials, although interestingly yields for individual reactions were as high as 99% demonstrating the varied reactivity that can be observed by reaction screening.  However, this work confirmed the difficulties observed for the C-N formation.

1536 Reaction Conditions in Parallel

 

When the reaction can be miniaturized and automated, the number of reactions possible is staggering.  Recently, the Merck Process Group reported their platform for high throughput experimentation for synthetic route scouting9.  Transitioning from microvials to 1536 well microtiter plates for 1 ul scale reactions, the nanomolar techniques represent the state of the art in the lower limits for testing chemical reactivity and reaction handling. For this work, they looked to the miniaturization of palladium cross-coupling reactions, including C-N bond formation, which as discussed above show difficulty on real-world substrates.  Focusing on finding a reaction that worked in DMSO and at room temperature, the group screened 16 second and third generation palladium pre-catalysts with 6 superbases for 96 unique reaction conditions. For a single bromide substrate with 16 nucleophiles and the 96 reaction conditions in one 1536 microtiter plate, using nanoliter dosing and rapid MISER HPLC-MS batch analysis the group was able to screen 1536 reactions in one day. This setup required only 0.02 mg material per reaction. Extension to additional substrates and reaction option led to a series of new nanoscale hits, with the validation reactions repeated at 1000x scale (~25 mg).  This HTE approach led to a new reaction protocol (or more accurately, a reaction screen) applicable to biologically relevant structures, where 21 complex products were synthesized at scale previously synthetically inaccessible. This work pushes the extent to which synthesis should not limit access to a molecule for probing a biological hypothesis.

As HTE tools further probe the synthetic toolkit, how best to merge the potential wealth of knowledge available from these systems for applications available to the bench chemist?  Most chemists do not have access to the type of instrumentation driving HTE experiments, certainly not at the discovery stage. Even with access, the instrumentation may not be supported by discovery budgets.  Screening thousands of conditions for one compound, in one program is not likely to be cost effective. When the chemistry is amenable to automation, as discussed, tools exist to screen thousands of reactions.  However, the setup, use and availability of a tool to screen 1536 may not be appropriate to screen conditions for just one product.  While easy to setup to run 1536 reactions, may not be the right tool to test 16 or 32 reactions for a single optimization.

When selecting the best approach to resolve reactions that do not work in a first pass, the medicinal chemist is forced to ask the question, “How much time and effort should I put into trying to make this structure?”  Once it is determined that optimization or screening is necessary, what options exist? With unlimited time, budget and robotics, one could test every possible combination to find the optimal reaction to make a new structure, and even develop some new chemistry along the way. However, at the bench, practical limitations exist. While designing an optimization, the easiest solution may be to select a few additional catalysts off the shelf (based on what is readily available), or try an extra base or a different solvent. A common limitation is the question “How many things do I want to weigh today?  For air and water sensitive reactions, this is an added burden and a further detriment to undertaking a large screen.

figure_3

Tools for reaction optimization are certainly not new and include everything from low to high tech solutions.  Parallel sets of reaction vials set up one at a time in the hood, specialized reaction blocks like the optiblock™ or bench top liquid handling units all afford the medicinal chemist to opportunities to screen catalytic reactions. However, the missing feature in many cases is the chemistry to use in the setup. Automated tools may sit unused while the hood is full of flasks and vials due to a lack of chemistry to fill an extensive array or the time required to setup the apparatus.

Finding the appropriate chemistry for a screen can be difficult. For example googling “Suzuki coupling catalysts” or turning to Scifinder™ to determine the best condition to fit each attempted coupling will reveal more than 50 commercially available palladium catalysts and thousands of literature references to decipher.    Several generations of palladium pre-catalysts have been development to accommodate a wide range of ligands to improve stability, reactivity, turnover and functional group tolerance10. How long does it take to select, order and test the new catalyst? Is 1 g of a new palladium catalyst necessary when only a few mg are necessary to test an unknown reaction?  Newly available catalysts and protocols are limited by the barrier between the literature and medicinal chemistry applications and require vetting.  Here the lessons learned from HTE studies can be informative.

HTE techniques to vet catalysts and reaction protocols for real world compounds in early discovery can inform medicinal chemistry programs even without access to the most technologically advanced chemical screening techniques. Data from HTE experiments like the Merck chemical informer libraries and other methodology studies represent opportunity to establish pre-designed arrays of chemical catalysts to be part of the standard medicinal chemist’s workflow. Standardizing knowledge learned from the many HTE experiments can be an opportunity to delivery chemistry to the medicinal chemist in easy to use formats increasing compound output and complexity without the need for access to HTE technology.

Pre-designed arrays of chemical reactions

 

Pre-designed catalyst kits represent a significant opportunity to both enhance discovery efforts by increasing the complexity of targets while saving time and money in the process.  Pre-weighed reagents and catalysts packed under inert atmosphere facilitate screening of reaction conditions quickly without the need to use a glove-box to manipulate air sensitive or moisture sensitive reagents.  Kits are prevalent in nearly all biochemical and molecular biology labs from the smallest academic to the largest of pharma. The use of kits in chemistry; however, remains rare even though reaction optimization is highly amenable to standardization.  Catalyst and reagent screening kits with pre-designed reaction conditions for the bench chemist can fill the gap between haphazardly screening a few catalysts or reagents that can be readily found near one’s own benchtop and extensive and time consuming literature searches for screen design.

For chemistry applications, “kits” has often taken on an unfortunate description.  Catalyst “kits” available from many suppliers may only contain 100 mg or more of each of a wide array of catalysts.  While convenient for ordering purposes, still require significant setup of the reaction and are often a limiting factor towards the utility of the full set of catalysts. To be truly beneficial, a kit should include all of the necessary components to run the reaction including reaction vessel.  Fully designed individual reactions overcome this burden of weighing and handling air and water sensitive reagents.  Catalyst screening kits for Suzuki-Miyaura, Buchwald-Hartwig amination and several other reactions are now commercially available11,12. These kit formats should be amenable to a broad range of chemistry.  New validated chemistry can be delivered to the bench chemist in easy to use formats quickly accommodating the adoption of new methodology without the burden of researching, designing and validating unknown chemistry.

 

Summary

 

Designing a reaction screen can be time consuming, inefficient and ultimately unsuccessful.   Weighting small quantities of catalysts under inert conditions is a large barrier to screening a wide array of conditions. As a result, the attempted screen will likely be compromised or hindered in some way, either in scope, scale or ultimately be given up entirely for higher priority or easier targets. Often after a few hurdles or setbacks, the structure that looked so important in the planning stages is discarded for expediency. However, each structured missed is an opportunity lost. Fewer and lower quality compounds synthesized results in less informed decisions when testing biological hypotheses. Screening kits with pre-designed reaction conditions for the bench chemist can fill the gap between haphazardly screening a few catalysts or reagents readily found near one’s own benchtop and extensive time consuming literature searches for screen design. Pre-designed catalyst kits represent a significant opportunity to both enhance discovery efforts by increasing the complexity of structure while saving time and money in the process.

Who we are:

HepatoChem: (a chemical technology company located in Beverly, MA for metabolite production, lead diversification of analogues and chemical kits for reaction screening.  HepatoChem specializes in developing catalyst platforms for late-stage functionalization and cross-coupling reactions with a focus on designing reaction arrays for specific functions. Two highlights include a biomimetic oxidation platform for metabolite production and late-stage fluorination.

What we offer:

EvoluChem Reagent/Catalyst kit systems enable handling arrays of air and water sensitive reagents without the need for a glove box or robotics. Kits solve a need for the bench chemist allowing quick screening of catalysts and reaction conditions to identify the fastest route for selecting a reaction of interest.  In addition to chemical kits for cross-coupling catalysis, HepatoChem offers kits for amide coupling, late-stage functionalization and biomimetic oxidation for metabolite production.

For further information, please contact info@hepatochem.com.

Request information about chemistry kits


Fields marked with an * are required








References:

(1) Quentin Michaudel, Yoshihiro Ishihara, and Phil S. Baran, “Academia–Industry Symbiosis in Organic Chemistry” Accounts of Chemical Research 2015, 48 (3), 712-721.

(2)  Megan H. Shaw, Jack Twilton, and David W. C. MacMillan, “Photoredox Catalysis in Organic Chemistry” J. Org. Chem. 2016, 81, 6898-6926

(3)  Tim Cernak, Kevin D. Dykstra, Sriram Tyagarajan, Petr Vachal, and Shane W. Krska, “The medicinal chemist’s toolbox for late stage functionalization of drug-like molecules” Chem. Soc. Rev. 2016, 45, 546-576.

(4) M. Garcia-Castro, S. Zimmermann, M. G. Sankar, K. Kumar, Angew. Chem. Int. Ed. 2016, 55, 7586.

(5)  (a.)  Miyaura, N.; Yanagi, T.; Suzuki, A. The palladium-catalyzed cross-coupling reaction of phenylboronic acid with haloarenes in the presence of bases. Synth. Commun. 1981, 11, 513−519. (b.) Paul, F.; Patt, J.; Hartwig, J. F. Palladium-catalyzed formation of carbon-nitrogen bonds. Reaction intermediates and catalyst improvements in the hetero cross-coupling of aryl halides and tin amides. J. Am. Chem. Soc. 1994, 116, 5969−5970.  (c) Guram, A. S.; Buchwald, S. L. Palladium-catalyzed aromatic aminations with in situ generated aminostannanes. J. Am. Chem. Soc. 1994, 116, 7901−7902.

(6)  For a comprehensive list of available cheminformatics software available, See  http://www.click2drug.org/

(7)  Artiom Cernijenko, Rune Risgaard, and Phil S. Baran, “11-Step Total Synthesis of (-)-Maoecrystal V” J. Am. Chem. Soc., 2016, 138, 9425-9428.

(8) We direct the reader to this excellent open access article:  Kutchukian, P. S.; Dropinski, J. F.; Dykstra, K. D.; Li, B.; DiRocco, D. A.; Streckfuss, E. C.; Campeau, L.-C.; Cernak, T.; Vachal, P.; Davies, I. W.; Krska, S. W.; Dreher, S. D. Chemistry informer libraries:  a chemoinformatics enabled approach to evaluate and advance synthetic methods.  Chem. Sci. 2016, 7, 2604− 2613.

(9)  Buitrago Santanilla, Alexander; Regalado, Erik L.; Pereira, Tony; Shevlin, Michael; Bateman, Kevin; Campeau, Louis-Charles; Schneeweis, Jonathan; Berritt, Simon; Shi, Zhi-Cai; Nantermet, Philippe; Liu, Yong; Helmy, Roy; Welch, Christopher J.; Vachal, Petr; Davies, Ian W.; Cernak, Tim; Dreher, Spencer D.  Nanomole-scale high-throughput chemistry for the synthesis of complex molecules, Science, 2014, 347, 49-53.

(10) Ruiz-Castillo, Paul and Buchwald, Stephen L. “Applications of Palladium-Catalyzed C-N Cross-Coupling Reactions” Chem. Rev., 2016, 116, 12564-12649.

(11). For a full catalogue of reagent and catalyst screening kits visit www.hepatochem.com.

(12)  KitalysisTM from Aldrich Millipore.