Applying automation in early drug discovery: lessons learnt and future perspectives

The low probability of success in discovering compounds coupled with the significant length of time it takes to ascertain clinical efficacy makes drug discovery an expensive and risky endeavour². The typical early drug discovery process can be compartmentalised into target identification, its subsequent validation by in-vivo and in-vitro testing, identification of compounds that can modify its activity from a variety of screening activities (lead identification) and its subsequent enhancement of the lead molecule that would result in a pre-clinical candidate molecule. The candidate would enter a period of efficacy, safety and absorption, distribution, metabolism, excretion and toxicity (ADME-Tox) testing in animals and ultimately enter the various human clinical studies³. It is only at the point in the process when the compounds are investigated in phase 2 human clinical trials that the efficacy of the compound in modifying the disease in the human population can be assessed. The time period for a drug discovery program to progress from its inception to a more mature state where human clinical trials are initiated is typically in excess of five years. A research program halted for efficacy, safety or other reasons after such a considerable length of time would have devoured sums of monies running into the region of tens of millions of dollars. Although these figures are dwarfed by the costs of late stage trials, with investments on the scale of hundreds of millions of dollars, the success of compounds in late clinical stages remains critically dependent on the quality of the outputs from drug discovery. The application of automation to improve the output from drug discovery has been one of the key themes within integrated pharmaceutical organisations in the past fifteen years. This article will discuss how appropriate automation in chemistry and biology discovery processes has become an important driver in improving efficiency and reducing timelines.

The screening of targets against compound libraries in a variety of assays in miniaturised microtitre plate formats, now including the commonly used 1536 well format, is termed high throughput screening (HTS). This approach which is considered to be relatively simple, low cost, rapid and highly efficient remains one of the most commonly utilised methodologies for identifying starting points for drug discovery efforts, especially in situations where directed drug design approaches are precluded by a lack of a protein structure or knowledge of the underlying biological mechanisms involved in the disease process. The small molecule screening collections of mid to large pharmaceutical companies are typically composed of 0.5 to 3.0 million distinct compounds in solution (usually DMSO) form and 10,000 to 100,000 available as solid form. The automation of the synthesis^4-8, purification^9,10, quality control¹¹, archiving and overall management of the compound library has been essential in ensuring that HTS campaigns have access to the chemical matter for primary screening and for subsequent re-testing. A pre-requisite for HTS is generation of a library of compounds. Although there are multiple organisations offering screening, library design and synthesis services on a fee for service or contract basis, many pharmaceutical companies remain biased towards maintaining their in-house library synthesis capability driven by a strategic requirement to preserve the exclusivity and novelty of their own compound collections. Automated processes are not confined to the parallel synthesis of primary screening collections, but are also applied to the hit follow up stage when new compounds are synthesised in order to establish target structure activity relationships (SAR) by exploring the chemical space around the hit series. In the early days of combinatorial chemistry, many compounds were of variable quality and purity and endowed with chemical properties that bore little relation to the desired profile of the final drug, as they contained undesirable features such as reactive groups or toxicophores. Adoption of chemical structure guidelines for screening collections, such as the five formulated by Chris Lipinski at Pfizer, and various refinements suggested by subsequent authors have lead to, arguably, the current generation of high quality compound libraries that contain potential starting points (lead-like molecule) for drug discovery with emphasis being placed on the compounds having low molecular weights and lipophilicity, thereby allowing room for their optimisation into a more drug-like molecule. These collections require expert stewardship to ensure that they can be delivered to the drug discovery work teams as and when they are needed. This not only a requirement for today’s current crop of targets but for the lifetime of the collections (typically ten to fifteen years) and high quality compound logistical operations will remain critical for long as compound screening remains a method of choice in early stage discovery. This has required the development of integrated facilities such that when compounds are synthesised or acquired, they can be placed into storage, typically in both solid form and as DMSO solutions of an appropriate concentration which would allow the determination of their potencies with respect to targets. The apparently rudimentary tasks of compound handling, manipulation including the weighing, dissolution and the complexities associated with this, which although appear trivial, are very time consuming for individuals. The automation of various processes around these functions has made the area of compound management one the most highly automated aspects of the entire pharmaceutical Research & Development process and the degree of human intervention has greatly decreased compared to the situation ten years ago. The physical storage conditions required to maintain the integrity of the compounds, also mitigate against the use of manual processes. Long term compound stores where the original stocks of compounds are at mM concentrations in DMSO are typically maintained at -20°C or below. With medium and short term storage in tube based formats suitable for cherry picking, many commercial systems have options for maintaining a N2 only environment. Maintaining a safe working environment and the requirement to minimise perturbations in the storage conditions (e.g. minimising unnecessary freeze-thaw cycles) have also driven the development of large scale automated compound storage and retrieval systems by many manufacturers.

The delivery of compounds to various assay scientists and movement of compounds in libraries in the size of 100,000s is not amenable manually, especially when taking into account these need to be screened against dozens of targets per year at various sites around the globe. Thus automation around the compound libraries and at the same time tracking compound location in assay plates is essentially a process that can not be executed reliably manually¹². There is also a need to ensure that the integrity of the compound collection is maintained and will necessitate the periodic analysis of samples using techniques such as on-line mass spectrometry.

One important future trend in HTS, which is especially dependent on robust automated compound handling processes, is the use of what has been termed iterative or directed screening. In this operating model, an organisation with a large screening library will select an initial subset (~5%) of all available compounds. Selection may be chemi/bio-informatics based, i.e. using the known structural features of the target or target class, or alternatively by an informed assessment of which compounds in the library are likely to be actives given their activities against related targets in previous screens. These compounds are then cherry picked from a direct access store and screened under medium throughput conditions. The hit population from these 5% are analysed with chemi-informatic tools and a further group (~ 1%) selected from the 95% of the file are selected for second round testing. Successive iterations of this process can then be use to identify the majority (>90%) of the potential hits contained within the screening file, after only screening a minority (<10%) of all possible compounds. The key dependencies are access to a large compound collection (>1 million) and an automated cherry picking facilities to obtain compounds quickly, preferably using a ‘single shot’ tube based format such as the REMP storage concept (Tecan, Switzerland). The realisable benefits of directed screening in terms of reduced reagent consumption and screening time can be substantial.

Removing process bottlenecks by automating the synthesis of compounds has also led to automation of downstream activities, especially in the screening (both low and high throughput) of compound libraries^13-16. In the early days of HTS, screening a biological target against a library containing large number of compounds was often done using mixtures of compounds (up to many tens of compounds per well), however, this requires considerable investment in time to identify which were the true actives in the mixtures. This has now almost entirely been replaced by screening using a single compound per test. In addition, manual prosecution of screening activities usually results in extended cycle times from the development of an assay to delivery of a validated hit. Efforts have been made in molecular biological and protein engineering techniques to allow the production of purified biological reagents in sufficient quantities to enable the screening of large numbers of compounds to be feasible. The development of microtitre plates of increasing density, from 96 to 384 and now the more commonly used 1536 density has allowed reaction volumes to be decreased from the region of 100µl to less than 10µl^17-20.

It is now also becoming commonplace to screen targets that were previously considered to be difficult to screen^21,22. In addition to assays using target centric approach there is also the System Biology approach that analyses the complete biological responses when screening and selecting hit compounds^23,24. In parallel with the development of higher density microtitre plates, reader technologies have been developed that allow the reading of a 1536 microtitre plate in less a minute. Similarly, liquid handling technologies now allow the accurate dispensing of a variety of reagents, including proteins, cells and scintillation proximity beads, the latter of which have proved to be particularly difficult to dispense. It is now almost a given to purchase reagent dispensing equipment that can dispense into the three main plate formats (96, 384 and 1536) and are capable of dispensing volumes in the low microlitre region.

Automation can also go beyond the HTS stage where selectivity and liability screening activities are carried out on the hits from the primary HTS campaign. If set criteria are met with respect to the potency and selectivity of compounds against the target, a lead optimisation phase can commence to alter specific characteristics of the molecule that could eventually lead to a candidate quality molecule. There has also been a drive to annotate compounds identified from screening campaigns with serum binding25, ADME26,27, solubility28, P450^29-34, clearance³⁵, physicochemical properties³⁶ and other key liability data^37-40 in order to progress the most promising ones. The conversion of a screening hit into a lead like molecule is an iterative process involving the synthesis of hundreds or thousands of compounds which are subsequently evaluated for their effect on the target. As more compounds are synthesised and assayed, a structure function relationship is built up and the results form the basis for the design of the next series of compounds. There have now also been significant advances in automation in mass spectroscopy^41,42 (including the processing, interpretation and management of the generated data), protein crystal harvesting, handling and image analysis^43,44. Strides are currently being made in calorimetry⁴⁵ and surface plasmon resonance⁴⁶ (providing stoichiometries of compounds and their targets in real time without the need for labelling compounds), NMR⁴⁷ (providing direct information on the affinity of the screening hits and the binding sites on proteins), microscopy⁴⁸ (allowing the study of protein function in their endogenous environment within cells) as these techniques are becoming more amenable to automation and increased throughput and will facilitate drug discovery.

Conclusion

The application of automation in early stage drug discovery offers an opportunity to achieve economies of scale and reductions in timescales whilst increasing quality and reproducibility. To be successful, however, the adoption of automation must serve the scientific requirements of the target rather than being a forced fit. This is especially true in the case of automated compound screening of bioassays where compromises on the biological relevance of assays should not be made in the cause of achieving the optimal level of process automation or miniaturisation. When properly applied, automation can add significant value to the overall process and allow de-risking of hit compounds and lead series by exposing them to the widest possible range of in-vitro liability assays. It is likely that bringing biology and chemistry closer together, coupled with the continuous advances in engineering and computation capacity will increase the number of opportunities where automation can be applied within by the pharmaceutical industry.

References

1. Booth, B. and Zemmel, R. Prospects for productivity. Nature Reviews Drug Discovery 2004, 3: 451-456.
2. Rawlins, M. D. Cutting the cost of drug development? Nature Reviews Drug Discovery, 2004, 3:360-364.
3. Welling P.G., Lasagna, L and Banakar, U.V. (eds.). The drug development process: Increasing efficiency and cost effectiveness. New York Marcel Dekker, Inc., 1996.
4. Hird, N. W. Automated synthesis: new tools for the organic chemist. Drug Discov. Today. 1999, 4:265-274.
5. Koppitz, M. and Eis, K. Automated medicinal chemistry. Drug Discov. Today. 2006, 11:561-8.
6. Edwards, P. J. The impact of parallel chemistry in drug discovery. IDrugs. 2006. 9:347-353.
7. Weber, A., von Roedern, E. and Stilz, H. U. SynCar: an approach to automated synthesis. J. Comb. Chem. 2005, 7:178-84.
8. Reader, J. C. Automation in medicinal chemistry. Curr. Top. Med. Chem. 2004, 4:671-686.
9. Ripka, W. C., Barker, G. and Krakover, J. High-throughput purification of compound libraries. Drug Discov. Today. 2001, 6:471-477.
10. Popa-Burke, I. G., Issakova, O., Arroway, J. D., Bernasconi, P., Chen, M., Coudurier, L., Galasinski, S., Jadhav, A. P., Janzen, W. P., Lagasca, D., Liu, D., Lewis, R. S., Mohney, R. P., Sepetov, N., Sparkman, D. A. and Hodge, C. N. Streamlined system for purifying and quantifying a diverse library of compounds and the effect of compound concentration measurements on the accurate interpretation of biological assay results. Anal. Chem. 2004, 76:7278-7287.
11. Squibb, A. W., Taylor, M. R., Parnas, B. L., Williams, G., Girdler, R., Waghorn, P., Wright, A. G. and Pullen, F. S. Application of parallel gradient high performance liquid chromatography with ultra-violet, evaporative light scattering and electrospray mass spectrometric detection for the quantitative quality control of the compound file to support pharmaceutical discovery. J. Chromatogr. A. 2008, 1189:101-108.
12. Keighley, W. W. and Wood, T. P. Compound library management. An overview of an automated system. Methods Mol. Biol. 2002, 190:129-152.
13. Houston, J. G., Banks, M. N., Binnie, A., Brenner, S., O’Connell, J., Petrillo E. W. Case study: impact of technology investment on lead discovery at Bristol-Myers Squibb, 1998-2006. Drug Discov. Today. 2008, 13:44-51
14. Liu, B. Li, S. and Hu, J. Technological advances in high-throughput screening. Am. J. Pharmacogenomics. 2004, 4:263-276.
15. Nettekoven, M. and Thomas, A. W. Accelerating drug discovery by integrative implementation of laboratory automation in the work flow. Curr Med Chem. 2002, 9:2179-2190.
16. Pereira, D. A. and Williams, J. A. Origin and evolution of high throughput screening. Br. J. Pharmacol. 2007, 152:53-61.
17. Gul, S., Sreedharan, S. K. & Brocklehurst, K. Enzyme Assays: Essential Data. 1998, John Wiley and Sons, Chichester & Bios-Scientific Publishers, Oxford.
18. Johnson, E.N., Shi X., Cassaday, J., Ferrer, M., Strulovici, B., Kunapuli, P.A. 1,536-well [(35)S]GTPgammaS scintillation proximity binding assay for ultra-high-throughput screening of an orphan galphai-coupled GPCR. Assay Drug Dev. Technol. 2008, 6:327-337.
19. Cassaday, J., Shah, T., Murray, J., O’Donnell, G.T., Kornienko, O., Strulovici, B., Ferrer, M. and Zuck, P. Miniaturization and automation of an ubiquitin ligase cascade enzyme-linked immunosorbent assay in 1,536-well format. Assay Drug Dev. Technol. 2007, 5:493-500.
20. Sorg, G., Schubert, H. D., Büttner, F. H. and Heilker, R. Automated high throughput screening for serine kinase inhibitors using a LEADseeker scintillation proximity assay in the 1536-well format. J. Biomol Screen. 2002, 7:11-19.
21. Dunlop, J., Bowlby, M., Peri, R., Tawa,G., LaRocque, J., Soloveva, V. and Morin, J. Ion channel screening. Comb. Chem. High. Throughput Screen. 2008, 11:514-522.
22. Dunlop, J., Bowlby, M., Peri, R., Vasilyev, D. and Arias, R. High-throughput electrophysiology: an emerging paradigm for ion-channel screening and physiology. Nat. Rev. Drug Discov. 2008, 7:358-368.
23. Butcher, E. C. Can cell systems biology rescue drug discovery? Nat. Rev. Drug Discov. 2005, 4:461-467.
24. Giuliano, K. A., Cheung, W. S., Curran, D. P., Day, B. W., Kassick, A. J., Lazo, J. S., Nelson, S. G., Shin, Y. and Taylor, D.L. Systems cell biology knowledge created from high content screening. Assay Drug. Dev. Technol. 2005, 3:501-514.
25. Hartmann, T., Schmitt, J., Röhring, C., Nimptsch, D., Nöller, J. and Mohr, C. ADME related profiling in 96 and 384 well plate format-a novel and robust HT-assay for the determination of lipophilicity and serum albumin binding. Curr. Drug Deliv. 2006, 3:181-192.
26. Whalen, K., Gobey, J. and Janiszewski, J. A centralized approach to tandem mass spectrometry method development for high-throughput ADME screening. Rapid Commun. Mass Spectrom. 2006, 20:1497-1503.
27. O’Connor, D. Automated sample preparation and LC-MS for high-throughput ADME quantification. Curr. Opin. Drug Discov. Devel. 2002, 5:52-58.
28. Alsenz, J., Meister, E. and Haenel, E. Development of a partially automated solubility screening (PASS) assay for early drug development. J. Pharm. Sci. 2007, 96:1748-1762.
29. McGinnity, D. F, Waters, N. J., Tucker, J. and Riley, R. J. Integrated in vitro analysis for the in vivo prediction of cytochrome P450-mediated drug-drug interactions. Drug Metab. Dispos. 2008, 36:1126-34.
30. Yao, M., Zhu, M., Sinz, M. W., Zhang, H., Humphreys, W. G., Rodrigues, A. D. and Dai, R. Development and full validation of six inhibition assays for five major cytochrome P450 enzymes in human liver microsomes using an automated 96-well microplate incubation format and LC-MS/MS analysis. J. Pharm. Biomed. Anal. 2007, 44:211-23.
31. Watanabe, A., Nakamura, K., Okudaira, N., Okazaki, O. and Sudo, K. Risk assessment for drug-drug interaction caused by metabolism-based inhibition of CYP3A using automated in vitro assay systems and its application in the early drug discovery process. Drug Metab. Dispos. 2007, 35:1232-8.
32. O’Donnell, C. J., Grime, K., Courtney, P., Slee, D. and Riley, R. J. The development of a cocktail CYP2B6, CYP2C8, and CYP3A5 inhibition assay and a preliminary assessment of utility in a drug discovery setting. Drug Metab. Dispos. 2007, 35:381-385.
33. Lim, H. K., Duczak, N. Jr, Brougham, L., Elliot, M., Patel, K. and Chan K. Automated screening with confirmation of mechanism-based inactivation of CYP3A4, CYP2C9, CYP2C19, CYP2D6, and CYP1A2 in pooled human liver microsomes. Drug Metab. Dispos. 2005, 33:1211-1219.
34. Jenkins, K. M., Angeles, R., Quintos, M. T., Xu, R., Kassel, D. B. and Rourick, R. A. Automated high throughput ADME assays for metabolic stability and cytochrome P450 inhibition profiling of combinatorial libraries. J. Pharm. Biomed. Anal. 2004, 34:989-1004.
35. Reddy, A., Heimbach, T., Freiwald, S., Smith, D., Winters, R., Michael, S., Surendran, N. and Cai, H. Validation of a semi-automated human hepatocyte assay for the determination and prediction of intrinsic clearance in discovery. J. Pharm. Biomed. Anal. 2005, 37:319-326.
36. Shaikh, S. A., Jain, T., Sandhu, G., Latha, N. and Jayaram, B. From drug target to leads-sketching a physicochemical pathway for lead molecule design in silico. Curr. Pharm. Des. 2007, 13:3454-3470.
37. Day, S. H., Mao, A., White, R., Schulz-Utermoehl, T., Miller, R. and Beconi, M. G. A semi-automated method for measuring the potential for protein covalent binding in drug discovery. J. Pharmacol. Toxicol. Methods. 2005, 52:278-285.
38. Tao, H., Santa, Ana. D, Guia, A., Huang, M., Ligutti, J., Walker, G., Sithiphong, K., Chan, F., Guoliang, T., Zozulya, Z., Saya, S., Phimmachack, R., Sie, C., Yuan, J., Wu, L., Xu, J. and Ghetti, A. Automated tight seal electrophysiology for assessing the potential hERG liability of pharmaceutical compounds. Assay Drug. Dev. Technol. 2004, 2:497-506.
39. Pohjala, L., Tammela, P., Samanta, S. K., Yli-Kauhaluoma, J. and Vuorela, P. Assessing the data quality in predictive toxicology using a panel of cell lines and cytotoxicity assays. Anal. Biochem. 2007, 362:221-228
40. Tong, X. S., Xu, S., Zheng, S., Pivnichny, J. V., Martin, J. and Dufresne, C. High throughput metabolic stability screen for lead optimization in drug discovery. J. Chromatogr. B Analyt. Technol. Biomed. Life. Sci. 2006, 833:165-173.
41. Morand, K. L., Burt, T. M., Regg, B. T. and Tirey, D. A. Advances in high-throughput mass spectrometry. Curr. Opin. Drug Discov. Devel. 2001, 4:729-735.
42. Chovan, L. E., Black-Schaefer, C., Dandliker, P. J. and Lau, Y. Y. Automatic mass spectrometry method development for drug discovery: application in metabolic stability assays. Rapid Commun. Mass Spectrom. 2004, 18:3105-3112.
43. Mooij, W. T., Hartshorn, M. J., Tickle, I. J., Sharff, A. J., Verdonk, M. L. and Jhoti, H. Automated protein-ligand crystallography for structure-based drug design. Chem. Med. Chem. 2006, 1:827-838.
44. Villaseñor, A., Sha, M., Thana, P. and Browner, M. Fast drops: a high-throughput approach for setting up protein crystal screens. Biotechniques. 2002, 32:188-189.
45. Weber, P. C. and Salemme, F. R. Applications of calorimetric methods to drug discovery and the study of protein interactions. Curr. Opin. Struct. Biol. 2003, 13:115-121.
46. Neumann, T., Junker, H. D., Schmidt, K. and Sekul, R. SPR-based fragment screening: advantages and applications. Curr. Top. Med. Chem. 2007, 7:1630-1642.
47. Damberg, C. S., Orekhov, V. Y. and Billeter, M. Automated analysis of large sets of heteronuclear correlation spectra in NMR-based drug discovery. J. Med. Chem. 2002, 45:5649-5654.
48. Starkuviene, V. and Pepperkok, R. The potential of high-content high-throughput microscopy in drug discovery. Br. J. Pharmacol. 2007, 152:62-71.