The impact of combinatorial chemistry on basic research and drug discovery
Dr. R. Houghten (Torrey Pines Institute for Molecular Studies, USA)
This was an opening talk designed to set the scene rather than report recent research results. It is not easy to summarise combinatorial chemistry in one hour, especially for an audience that contains so many real experts. Houghten did so with considerable skill, keeping the attention of the audience, but with an inevitable slight bias towards tea-bag technology, mixtures, the numbers game, positional scanning and libraries from libraries.
Designing an epitope display system on filamentous bacteriophage; epitope structure and vaccine design
Dr. R. Perham (University of Cambridge, UK)
I now understand a lot more about the intricacies of phage display than I did before I heard this talk. Perham is Professor of Structural Biochemistry at the University of Cambridge and his patented technology is employed by Peptide Therapeutics.
Combinatorial libraries for studying molecular recognition
Professor G. Lowe (University of Oxford, UK)
This was an excellent talk on which I made copious notes. Lowe first discussed how base pair sequences are recognised and whether there is a code that can be cracked. It turns out that the beta-sheet is a highly effective recognition motif and the convex surface is more significant than the concave one because it is able to penetrate much deeper and more successfully.
The natural peptide antibiotic tachyplesin I is an example of a folded polypeptide chain stabilised by two pairs of cysteine bridges. Lowe has made a similar molecule with D-proline and N-methyl alanine included because they are known to encourage beta-turn formation. The disulphide bridges form spontaneously and S T K D-P S T K are on the convex face. The bridges were studied with 2D NMR.
Having found a beta sheet structural probe, they now needed a DNA library. They made a selection template which contains a central 14 base pair random combinatorial library flanked by sequences which are recognised by Type IIS and Type II restriction enzymes. Lowe concluded by describing a protection, selection and amplification cycle.
Automated parallel synthesis of small organic molecules on solid phase
Dr. M. Lebl (Trega Biosciences, USA)
Lebl discussed why there is a need for robotic synthesis, what is available, Trega's approach, and the components of automation. He concluded with an example of synthesis. Being a Czech himself, Lebl likes to remind us that the word "robot" was coined by a Czech playwright.
Synthesisers are available from Advanced ChemTech, Argonaut, Tecan, Bohdan Automation and Rapp Polymere. Alternatives for robotic platforms are Zymark, CRS, Sagian/ORCA and Yaskawa. Trega has made 2.3 million heterocyclic and acyclic compounds, from 20 templates, as individual compounds or as pools. The Trega auto-synthesiser uses a modular approach with independent and disposable reaction vessels. A combination of automation and manual methods is used. They need to make 25,000 compounds a week. They use a combination of tea-bag and semi-open deep well plate reaction vessels.
They have software for automatic generation of a compound database, for eliminating undesirable compounds and for molecular diversity (Tripos and MSI). They use a Yaskawa robot with 3 robotic heads, which handles 72 plates at a time; a Quadra 96 for making 4 plates from a master; Packard Canberra multiprobes; a Gyromax 737 Amerex heated shaker; a Savant SpeedVac; and gaseous HF cleavage towers. Lebl showed a scheme for "hubotic" synthesis: this uses both humans and robots. He showed photographs to illustrate how the procedure works and briefly discussed analytical quality control.
New approaches to solid phase syntheses and screening
Professor K. Burgess (University of Texas A & M, USA)
Burgess' group made oligoureas by split synthesis and needed a photolabile linker as part of the process of identifying products by MS. Use of TentaGel and synthesis of a photocleavable amino acid proved not as easy as they had hoped but eventually they made a heavy one incorporating a para-bromophenylsuplhonyl group. A single bead was put in a MALDI probe and the bromine atom helped to distinguish real peaks from noise in the MS.
Turning to a second topic, Burgess talked about some work to see how much faster the OF 4949 series of cyclic peptides could be made by solid-phase cyclisation via an SNAr reaction rather than in solution. Sixteen- and seventeen-member ring derivatives were also tried - use of the milder base TBAF, rather than potassium carbonate helped prevent epimerisation. 19-F NMR was used in characterisation. Kevin's Holy Grail is the synthesis of vancomycin and he discussed the formation of the C-O-D-O-E ring skeleton.
Finally he talked about high throughput screening of libraries of catalysts. Arrays for catalysis were synthesised in a glove box in an inert atmosphere. Mitosenes were made by a symmetric C-H insertion reaction. The usefulness of AgSbF6 might not have been discovered without HTS.
Efficiency by design. Optimisation and automation in process research
Mr. T. Owen (Glaxo Welcome, UK)
Experimental design is a statistical method of handling data. It "takes the art out of science and replaces it with science". Owen described a case study in which the conversion of a silyl ester to an hydroxy ester was producing a lactone impurity. The object is not simply to maximise the yield of the main product. It is important to control the impurity to a specified level, maximise throughput (time and concentration), minimise waste and material costs, and obtain a robust process, which gives consistent yield and purity.
HPLC curves show that the highest yield is after 9 hours but the amount of lactone impurity increases such that 6 hours is a better compromise. It is possible to vary both time and temperature : Owen showed a two factor response surface and a contour map. A set of experiments can be designed where all the factors are simultaneously altered. Owen illustrated 4 factor design (time, temperature, amount of reagent and amount of solvent). Multiple experiments were carried out and one result was predicted. The statistical package DX5 can produce perturbation plots. The goal was to maximise product, minimise impurity and minimise unchanged starting material. The software allows factors to be weighted but chemical considerations are also important: for example, the silyl ester starting material can be easily washed away whereas it is hard to get rid of the lactone impurity. The effects of process deviation can be modelled. It was then necessary to see if the model was valid after scale-up.
Experimental design is a powerful technique but it is repetitive and daunting to apply. This is where automation comes in. Owen showed the Development Automated Reaction Toolkit (DART): a Gilson 233XL autosampler with customised Stem reaction stations.
A combinatorial selection approach to catalytic receptors
Professor J. K. M. Sanders (University of Cambridge, UK)
The approach to combinatorial chemistry here is reversed: combinatorial chemistry is used to create new receptors. A library of building blocks is set up. Each building block will be equipped with recognition sites and unique mass or spectroscopic reporters. A transition state analogue with a tether selects from the cocktail those components that bind it best. The selection is carried out under thermodynamic, equilibrating conditions so that incorrect bond-making will be proof-read and on average rejected. The final solution may contain millions of different species but if the transition state analogue is tethered to a bead it will be readily isolated from the mixture together with its ideal host.
Transesterification was a useful reaction to study and work on various macrocycles has already been published. Certain structural components proved more useful than others for MS, electronic and NMR spectroscopic techniques to be applied. Unpublished results of Stuart Rowan's, demonstrating reversibility and the production of mixtures were presented.
Combinatorial approaches to selective catalysis
Professor E. N. Jacobsen (Harvard University, USA)
Combinatorial catalysis is not an ideal term to use but the technique has utility for discovery of new catalysts for important reactions (in lead discovery); for optimisation of catalyst selectivity and reactivity (for lead optimisation); and for selective catalysis in combinatorial synthesis (new methods). Jacobsen showed 3D structures of template bound cyclic peptides bearing RGD. Enantiomeric templates induce different beta-turns.
The next part of the talk concerned privileged chiral catalyst structures. Jacobsen has studied aziridine asymmetric ring opening and has made a parallel library of tridentate Schiff bases in autosampler vials. He has also worked on a ligand/metal library catalysed Strecker reaction.
Finally he described a combinatorial approach to selective catalysis. Two amino acids were separated by a turn element that should create a metal binding environment. Metal ion complexation was studied by bead colouring. About 20 metals, particularly lanthanides were tried.
SELEX: drug candidates in a hurry
Dr. L. Gold (NeXstar Pharmaceuticals, USA)
The first day's papers were followed by a poster session, which was followed by dinner and Gold's paper was given after dinner. Despite the length of the day, most delegates turned up to the talk and they were rewarded with a dazzling presentation. Perhaps it was a sales pitch, perhaps the science has snags but Gold can certainly make it all credible. This man has personality and knows how to keep the attention of an audience. He talked about "aptamers", chemically modified oligonucleotides. They have the disadvantages of high molecular weight and the need for parenteral administration but the SELEX method does repeatedly give high affinity drug candidates. It seems that this is because the shape of the oligonucleotide produces a large interface between the drug candidate and the target. The focus is on extra-cellular targets. The method aims to produce drugs, not leads. Aptamers are not immunogenic or toxic. The Product Anchored Sequential Synthesis (PASS) process is inexpensive and scaleable.
Combinatorial chemistry and cheminformatics
Dr. A. Polinsky (Alanex Corporation, USA)
Polinsky likened throughput in target discovery, lead discovery and the lead-to-drug to interconnecting pipes. In the old days all three pipes had approximately the same "bandwidth". Now molecular biology has been added to biochemistry as a target discovery tool, and screening is much faster but medicinal chemistry throughput has not increased. In the future, genomics will produce yet more targets, there will be ultra HTS and combinatorial chemistry and medicinal chemistry will be comparatively even slower in comparison. This is where the bottleneck is and Alanex is addressing the problem with combinatorial medicinal chemistry and cheminformatics.
For combinatorial medicinal chemistry they have a toolbox of validated combinatorial reactions and custom building blocks. Polinsky spoke about high throughput precursor synthesis. Another philosophy is that it is economically unwise to concentrate on making an exploratory library 85% pure.
Cheminformatics encompasses chemical database technology, QSAR, molecular modelling and diversity calculations. Alanex's LiBrain software teaches computers to make decisions. The computer is given knowledge of synthetic chemistry. By perception of chemical structures and SAR it can choose descriptors, do affinity element analysis, and handle diversity, similarity and QSAR. As yet, though, knowledge of ADME properties is an unsolved problem and avoiding patented structures is something about which LiBrain has no ideas.
Fluorescent beads, assays and substrate specificity
Dr. S. Balasubramanian (University of Cambridge, UK)
Balasubramanian has worked with fluorescent molecules attached to beads. He discussed complications that relate to quenching effects and band shifts that depend on the loading. Anthracene, Texas red and Dansyl were some of the fluorophores studied as tags. He has been exploring combinatorial approaches to studying enzymes involved in protein phosphorylation and has developed a novel on-bead approach to screening a phosphotyrosine peptide library against a protein tyrosine phosphatase to select substrates. He has used Kieselguhr, PEGA, Toyopearl and TentaGel resins but says that there is a need for enzyme compatible solid phases to be developed.
No static at all: using radio frequency memory tubes without (human) interference
Dr. B. Prom (IRORI, USA)
IRORI is a pioneer in radio frequency (RF) tags for combinatorial libraries. Their AccuTag-100 system purportedly combines the advantages of both parallel synthesis and split and pool. The system combines glass encapsulated microchip tags, polymer microreactors and software, to produce a turn-key solution for combinatorial chemistry.
Radiofrequency encoded combinatorial chemistry makes discretes in multimilligram quantities and uses non-invasive labelling. The process is called "directed sorting". The technology involves chemically inert microreactors, miniature RF tags, a benchtop scanning station and synthesis management software. "MicroKans" or "MicroTubes" may be used.
Structure-activity and information from ECLiPS™-encoded combinatorial libraries
Dr. E. McDonald (Pharmacopeia, USA)
The binary encoding strategy of Clark Still and Mark Wigler, using halo aromatic tags detected by EC-GC, is so well known that I will not detail it here. McDonald gave a very clear and enthusiastic description of the technology which has been adopted at Pharmacopeia for producing large, diverse, quality controlled libraries sold in microtitre plates. They have 2.8 million compounds using more than 300 scaffolds.
For solution assays they photolyse one third of a compound into solution and make master plates. Derivative plates are screened at 20 compounds per well. Single beads from an active well are rephotolysed for assay at 1 compound per well. McDonald discussed results from two types of G protein coupled receptors in screens and showed what I call the "skyscraper diagrams" that everyone uses to illustrate the success of their assays. He also touched on diversity and appropriate coverage of DVS chemical space.
Towards combinatorial biosynthesis of complex polyketides
Dr. P. Leadlay (University of Cambridge, UK)
This was such an excellent talk that I am at loss as to how to do it justice in a paragraph or two, especially since I am not a specialist in polyketide chain building. Some people felt that this was one of the best papers at the conference. The research aims to throw light on the fundamental rules governing the chemistry and stereochemistry of chain growth in multienzymes.
A combinatorial blending of chemistry and biology
Professor K. Janda (The Scripps Research Institute, USA)
Janda's theme was mainly on the chemistry of PEG and PEG-like polymers as soluble supports for catalysts, reagents and combinatorial operations. Unfortunately I was not in my usual position at the front of the lecture theatre for this presentation and I had considerable trouble taking my usual reams of notes because of the amount of material on Janda's slides and the unusually large numbers of creative ideas he manages to pack into a presentation. Fortunately, he does publish his results regularly in significant journals.
Ligand design for affinity chromatography of biopharmaceuticals by intelligent combinatorial chemistry
Dr. C. Lowe (University of Cambridge, UK)
Refer to abstract.
Making antibody and peptide ligands by repertoire selection technologies
Dr. G. Winter (University of Cambridge, UK)
Refer to abstract.
Advances in the design, synthesis and evaluation of small molecule libraries
Professor J. Ellman (University of California, Berkley, USA)
This paper is difficult to cover well here because it contained masses of excellent chemistry which really demands structural diagrams and reaction schemes. I cannot assume that all readers will have state-of-the-art browsers, applets and chemistry plug-ins. So, I can only give a flavour of what was obviously one of the key presentations that attracted people to attend the meeting.
Ellman has been using palladium mediated approaches to prostaglandin synthesis. Lorin Thompson and Young-Choon Moon have been doing some of this work, for example, introducing diversity in COOR at the end of the longest substituent chain. Ellman also talked about some of his work on beta-turn mimetics. Alex Virgilio and Stephan Schurer have found 58-363 nanomolar leads in a focused library of 176 mimetics in assays using somatostatin.
Renin, pepsin, gastricin, HIV protease, cathepsin D and E and plasmepsin are all aspartic acid proteases. Ellman and Kuntz have used structure-based design to select the functionality required in a hydroyethylamine to target, in particular, cathepsin. Ellen Kick and Jon Ellman published an hydroxyethylamine isostere library in 1995. Diana Roe docked a virtual library for inhibiting cathepsin using DOCK. Diverse libraries have been made by Jarvis Patrick clustering and Daylight fingerprints. Directed libraries were also made. From an optimisation library, 9, 14 and 15 nanomolar leads in a cathepsin D fluorescence assay have been found. In another project, Christina Lee and Ellen Kick developed a new linkage strategy for chelate controlled reduction.
New solid phase chemistry approaches to drug-like molecules
Dr. A. D. Baxter (Oxford Diversity, UK)
Baxter described a collaboration with Leukosite to develop selective inhibitors of MAdCAM which have potential as anti-inflammatory agents. Goals were to find a 1 micromolar lead, to study solid-phase synthesis methods and to use novel amino acid building blocks and templates. In designing peptidomimetics based on LDT they needed a good N cap, then a good C cap, and they wanted to use non-natural amino acids.
From the capping work they did make a 1 micromolar lead and the purities of the compounds were excellent. Then they went on to use the non-natural amino acids D-Leu, D-Asp and D-Thr and some beta amino acids. Baxter showed two routes to beta-(S)-leucine, one of them using Steve Davies chemistry. Some tetrazole building blocks were particularly interesting to the audience.
Oxford Diversity (OD) have also been doing a separate project with Pfizer on development of silicon based linkers. Selective cleavage under mild conditions was required and they wanted to recover the material. Amide coupling was preferable. The OD linker cleaves better and faster than one used by SmithKline Beecham.
Generation and screening of encoded combinatorial libraries
Dr. M. Gallop (Affymax, USA)
With binary encoding, relatively few tags encode lots of information. Affymax have used secondary amine tags in quantitative encoding. Decoding involves selecting a bead, stewing it in HCl, then dansylating the mixtures of amines. The dansylated tags are then analysed by HPLC. LC/MS of underivatised tags is also possible but there are some snags to the use of MS.
Gallop went on to talk about screening of encoded bead libraries. Single bead assays require high loading, specialised automation for bead handling and orthogonal chemistries for ligand and tag. Photolysis is used before solution transfer and detection. ACL 1534, a 3-mer amino acid library (the alpha-methyl nitroveratryl linker has been published) was screened against CCKa receptor. Gallop described tiered release and some 150-840 nanomolar leads. He also talked about some 1,4- and 1,5-benzodiazepine privileged structures and an SNAr strategy. He discussed some idiosyncrasies of the cyclisation stage.
A unique approach to drug discovery utilising information-rich libraries: an alternative to synthesising and screening large random compound libraries
Dr. P. Myers (CombiChem, USA)
CombiChem libraries are designed for information content. Single, pure compounds, in small libraries are made. The Universal Informer library contains 10,000 compounds. Myers distinguished lead optimisation, lead evolution and lead generation. A "good" library with high information content contains promiscuous compounds, i.e., those that are capable of interacting with many targets.
Myers described CombiChem's trademarked Discovery Engine and the use of hypotheses in a cycle that is iterated 5 or 6 times to convergence. It should take 2-3 months to get from hypotheses to compounds. The company believes in 3D descriptors and the handling of conformational flexibility in library design. A virtual library of 5000 billion molecules is possible but synthetically feasible molecules are required with 2-4 functional positions. The library is searched by hypotheses for daughter libraries.
The flexible and feature rich Universal Informer Library of promiscuous molecules is the first step in the discovery process but CombiChem also makes virtual libraries and daughter libraries. Myers also used a "skyscraper" diagram to prove that hits can be found.
Radical cyclisations and cycloadditions in parallel synthesis
Dr. R. Armstrong (Amgen, USA)
Armstrong described the automation techniques used to produce three libraries a week of 3200 compounds. There is a different workstation for each operation and humans move things from station to station. A 400 compound block is used in a Tecan workstation. Armstrong talked about resin capture for compounds related to raloxifen and tamoxifen. Only the desired product is captured by the resin. He also talked about samarium iodide mediated cyclisation, a 5-component Ugi reaction, an Ugi reaction with COS in place of OCO, and trialkylamines via Hofmann elimination.
Parallel synthesis of hydroxamic acids
Dr. M. Whittaker (British Biotech Pharmaceuticals)
Whittaker spoke about combinatorial methods for MMP inhibitors. His group has developed both solid phase and solution phase routes to hydroxamic acid compounds. The first part of his talk covered O-hydroxylamine functionalised polymers supports in the preparation of succinyl based compounds. Wang resin gives impurities on cleavage but 2-chlorotrityl-O-NH2 and HMPB-MBHA O-hydroxylamine are useful acid-sensitive resins for producing "left hand side" MMPIs, (Z-A1-A2- A3-NHOH).
For right hand side MMPIs (cf. batimastat and marimastat), they decided to try an Ugi 4-component reaction with Rink and PAL resins. They got good results with ammonia in place of Polymer-NH2. Clean-up involves ion exchange before the Ugi products are converted to hydroxamate MMPIs. They have made many succinates and have used automated reverse phase preparative HPLC. Results versus HFC, 72-KDa Gel and stromelysin-1 were given.
Strategy for the synthesis and screening of small molecule combinatorial libraries
Dr. S. Rahman (SmithKline Beecham Pharmaceuticals, UK)
Rahman has used a bead associated screening strategy for combinatorial libraries, arraying single beads into individual wells, releasing the compounds into solution and reconstituting them in target compatible solvents for assay. With this method, large mixtures can be screened as individual compounds. Unique molecular weights are needed for MS and SB has developed a proprietary method for avoiding mass redundancy. [I saw a poster about this at a CHI meeting earlier this year.]
High capacity, large beads (250-300 micron) are needed. Rahman spent some time describing quality assurance. A 100 compound polymer bound lysine library was made, and the beads were arrayed into plates. Cleavage was done in three ways to introduce a third point of diversity. Leads to a novel target were identified.
Combinatorial chemistry: a perspective
Dr. D. Brown (Glaxo Welcome, UK)
The focus at Glaxo Wellcome is on quality and information. Diversity libraries are used where there is zero target knowledge and diversity is important. Focused libraries are used in the protein class case; pharmacophore based libraries work if a pharmacophore is known; and structure based design is used if protein X-ray has given a known target structure.
Random, drug-like libraries involve expanding the corporate compound collection by suitable acquisitions, and enumerating virtual libraries and choosing diverse sets using, for example, genetic algorithms (GA). John Bradshaw has developed TRIAGE software, based on the Daylight toolkit, for set selection, library comparison and compound selection for screening. Darren Green and Peter Willett (Sheffield University) have developed a program, RECAP, using a GA for monomer selection and focused library design. (Since the meeting I have had a message from Darren Green saying that RECAP is Xiao Qing Lewell's work for identifying monomers targeted for certain protein classes whereas Green has been involved with using a GA to design diverse 'drug-like' libraries with Peter Willett and Val Gillet at Sheffield. Peter Willett has also sent me a message asking me to inclue Val Gillet's name since she was principally responsible for this elegant piece of work).Pharmacophore based libraries can be developed by traditional medicinal chemistry skills.
The second part of Brown's talk concerned parallel synthesis techniques for lead optimisation. A range of innovative methods has been developed to speed up solution phase synthesis. Batches of 20 or 80 compounds are made via Stem-reflux or Stem-cool stirrer hotplates or MTP blocks. Purification methods include a simple "lollipop" technique, membrane technology to separate aqueous and organic phases, the use of resin based scavenging agents, parallel centrifugation, parallel solvent blow-down and parallel cartridge based chromatography. Andy Merritt and Dick Storer were credited with some of this work.
Open access LC-MS is used for quality control. Brown also mentioned the use of 23 assays for lead optimisation. There are Tecan Genesis robots in the enzyme laboratory and various automation techniques are used. Efficient data handling is essential - Activity Base and the Glaxo Web were mentioned. There has been a 10-fold increase in assay productivity since 1995. Mike Snowden, Graham Baker and Mike Malloy were mentioned. Manual operations before and after synthesis are new bottlenecks. Brown concluded with some comments on implications for the future: the changing role of laboratory personnel, changes in laboratory design and changing traditional medicinal chemistry department structures.