High-throughput Technologies in Drug Discovery

Bringing new medicines to patients struggling with serious illness is the driving force behind drug discovery and development. Originating as small molecules, drugs have shifted to large molecule therapies and will continue to shift as precision medicine is adopted. In spite of these different types of drugs, the development time to create them has not changed dramatically. It still takes 10 to 15 years to bring a drug to market. For this reason, there is tremendous interest when it comes to finding ways to develop drugs with higher throughput methods. This article highlights advances in three technology areas that show great promise to speed the drug discovery and development process.

Automation and robotics

The need for automation and robotics is not a new topic in drug discovery but an area where great progress has been and continues to be made. Automation is essential to implementing high-throughput strategies. What started out as a solution to achieve higher throughput has extra benefits. An automated process provides better data quality due to process consistency. Human error is minimized, and the presence of an audit trail allows traceability if questions arise. Automation also provides walk away freedom for the scientist to pursue other tasks.

To witness an automated set up, one may find that each automation step may not be fast in and of itself. But consider the shift away from an 8-hour employee workday to one of continuous 24-hour operation. Screening projects are now reduced by at least a factor of 3, thus yielding higher throughput.

Automation can be categorized into three general modes defined as (1) batch, (2) semi-automated, and (3) integrated. The three modes range from limited to extensive on key automation criteria which include things like flexibility, walk away capabilities, number and complexity of tasks. Batch mode, for example still requires a scientist to load stacks of plates that then are subjected to a limited step in the process. Integrated automation, the most sophisticated, is capable of carrying out multiple scheduled steps facilitated by a robotic mover. This allows unmanned operation for extended periods providing walk away or overnight convenience.¹

An important consideration in any automated solution is the skill requirement of the operator. More sophisticated systems will require automation programming skills often leveraging automation engineers. Specialized training from the equipment vendor may also be required. Batch automation can often be accomplished with little specialized training. Compared to ten years ago, automation today has evolved and become more democratized. This continued trend will reduce the need for specialized training in the future with more turn-key and intuitive solutions becoming commercially available.

David Ebner, Principal investigator at the Target Discovery Institute in Oxford, UK explains: “There are two key limiting factors for any group trying to do high-throughput screening today.  The first is the expertise to translate a benchtop assay to a high-throughput platform and the second factor is the expense.”  Centralized core facilities are one way to address the expense factor. And as automation becomes more turn-key, the need for specialized engineers is reduced, allowing future resources to be more research scientist driven.

Miniaturization

Most steps of high-throughput lead discovery have been affected by miniaturization and parallelization.² Increasing plate well density beyond a 96-well standard format is the first step in miniaturization for automated assays. Target densities of 384, 1536, and even 3854 wells per plate are available. Not only is higher throughput screening achieved through this miniaturization, but reagent costs are reduced as reaction volumes go from 10–20 mL in the well of a 384-well plate down to <2 mL in the well of a 1536-well plate.³

Fluid handling in miniaturized assays can be quite challenging, however it is crucial for performance.  Fast, accurate, and controllable dispensing is especially difficult when compounds are stored or solubilized in organic solvents. Effective mixing, evaporation and clogging are additional problems to overcome.^{3, 4} For miniaturized cell-based assays Ebner adds that: “One of the most common problems that high-throughput labs have to address is spatial or edge effects.”  Edge effects are known artifacts, typically found in perimeter wells of a microplate, that experience poor cellular growth compared to cells in the rest of the plate. All of these challenges tend to force the practical plate density to 384-wells.

Microfluidic technology, a more extreme form of miniaturization, addresses some of these known fluid handling challenges.⁴ Microfluidic chips provide the benefits of reduced volumes while replacing liquid handling mechanics with channels connected to liquid reservoirs. In some cases, the device has integrated tools such as electrodes built-in and can combine multiple operational steps. ⁵

Microfluidic devices are also able to isolate single cells, which can be further cultured on the chip. This ability removes cellular heterogeneity on cancer cell populations as an example. Traditional drug screening methods see response information from an average of all cells. The microfluidic solution allows analysis of a single cell's antidrug response.⁴ In addition to this cell-on-chip model, recent advances have led to tissue-on-chip and organ-on-chip models which are still early in development. These kind of chip models may someday provide a powerful alternative to animal models.⁶ Because they are early in development, they are not high- throughput solutions today. But they show great promise to speed determination of drug activity, optimal combinatorial drug screening and toxicity testing in the future.⁴

Artificial intelligence

Applied to drug discovery, artificial intelligence (AI) has been used in medicinal chemistry for designing compounds since the 1960s. ⁷ Machine-learning tools like quantitative structure-activity relationship (QSAR) modeling have identified potential target molecules from millions of candidate compounds. ⁸ Today, AI has expanded its application in drug discovery to a range of tasks from robotics control to image analysis and logistics. AI has also been applied throughout the drug discovery process from target selection, hit identification, lead optimization through to preclinical studies and clinical trials. ^{7, 8}

Dr Mohammad HamediRad and colleagues at the University of Illinois explain that with new uses of AI, "the role of researcher changes from drivers of the experiments to supervisors of the system." AI, integrated with robotic systems enables automation of the design, build, test, and learn (DBTL) cycle. This results in a platform that designs experiments, executes them, analyzes the data then optimizes and executes subsequent experiments iteratively. This closed loop discovery reduces the total number of experiments and generates the best possible optimization. The concept was demonstrated by HamediRad and colleagues in 2019. Their fully-automated platform evaluated less than 1% of possible variants and outperformed traditional screening methods by 77%.⁹

AI platforms can cut down the development time from lead molecule to a candidate by more than half. AI predicted molecules are more likely to be correct and allow a focused effort.  Time isn’t wasted testing irrelevant molecules which would have been worked on otherwise and make up 90% of the molecules tested by traditional methods.¹⁰ “Currently, AI can help find novel compounds which are more potent and selective using high quality screening data sets much faster and at less expense than screening alone,” explains Ebner.

Personalized or “precision” medicine is another area where AI plays an important role. Precision medicines are a growing proportion of drugs in the industry pipeline. ¹¹ Extensive collections of human samples (diseased and healthy) are required for biomarker identification in developing a personalized medicine. ¹² Typically, all samples are sequenced using next-generation sequencing which generates massive amounts of data. AI methods of deep-learning make analysis of these big data sets possible. ⁸

Automation Journey Guide: How To Automate Simple to Complex Workflows For Achieving Results Beyond High-throughput

Laboratory automation is playing a key role in advancing scientific research, from pharmaceutical development to diagnostics. Whether it is to automate a simple or sophisticated workflow, automation is now used in labs throughout the world to increase their capacity and throughput. In this eBook, discover a detailed guide to introducing automation to your lab.

Download eBook

Sponsored Content

Bringing drugs to market faster

The path to developing a drug is a long one. Advances in technology will affect the speed the way drugs are developed. Here we reviewed the current state of laboratory automation, miniaturization and artificial intelligence. Innovation will continue on all of these fronts. Automation and robotics will become turn-key and democratized, microfluidics will evolve chip models that may reduce animal testing, and artificial intelligence may change the role of the drug discovery scientist. The combination of these advances and the development of new ones will make sure the best drugs are brought to market with increasing speed.

References

1. MJ. Wildey et al. (2017) High-throughput screening. Annual Reports in Medicinal Chemistry. 50, 149-195.
2. D. Cronk. (2013) High-throughput screening, pages 95-117. Drug Discovery and Development (Second edition).
3. D. Dunn and I. Feygin. (2000) Challenges and solutions to ultra-high-throughput screening assay miniaturization: submicroliter fluid handling. Drug Discovery Today. 5(12), S84-S91.
4. J. Sun et al. (2019) Recent advances in microfluidics for drug screening, Biomicrofluidics., 13, 061503.
5. P. Dittrich and A. Manz. (2006) Lab-on-a-chip: microfluidics in drug discovery. Nat Rev Drug Discov, 5(3), 210-218.
6. C. Probst et al. (2018) High-throughput organ-on-chip systems: Current status and remaining challenges. Current Opinion in Biomedical Engineering. 6, 33-41.
7. M. Sellwood et al. (2018) Artificial intelligence in drug discovery. Future Med Chem. 10(17), 2024-2028.
8. Zhang et al. (2017) From machine learning to deep learning: progress in machine intelligence for rational drug discovery. Drug Discovery Today, 22(11), 1680-1685.
9. M. HamediRad et al. (2019) Towards a fully automated algorithm driven platform for biosystems design. Nat Commun. 10, 5150.
10. SLAS (2019) AI and machine learning facilitates faster path to personalized medicine. Available at: www.SLAS.org.
11. A.Qu and T. Shuster. (2019) The future of drug development: the implementation of precision medicine for successful oncology drug development. Available at: www.pharmavoice.com.
12.  J. Pandya (2019). Biobanking is changing the world. Available at: www.Forbes.com.