Drug discovery is the first step towards the development of new therapeutics. From the first screening to the launch of preclinical testing the process is long, coupled with high cost and uncertainty, how can artificial intelligence technologies help to accelerate this process and make it for effective?
by Deborah Seah
Finding new medicines for treatment of disease involves many years of research and development into the product before it can finally be used for commercial consumption. The cost of drug discovery and development is estimated to be around US$ 2.6 billion. It would also require almost 12 years of work by scientists and pharmaceutical companies for a new medicine to be taken from drug discovery to finally be launched for commercial consumption. An estimated one-third of the cost is spent in the drug discovery stage. On top of that there is a high failure rate of 90 percent, with many molecules not making it past phase I clinical trials. Improvements in drug discovery for the future of drug development are necessary to enhance efficiency and produce cost-effective methods in delivering new therapeutics to the market.
Pharmaceutical companies are now tapping in to use artificial intelligence (AI) technologies to help in the process of drug discovery and development. AI is an umbrella term used to describe the ability of machines in copying the cognitive and information processing capabilities of the human mind.
AI technologies can be further branched out into the following:
- Machine Learning: This application of AI uses algorithms in data analysis and the building of automated analytical model building. It enables the system to learn by itself from structure and unstructured data to make changes and identify patterns based on experiences without the need for external intervention.
- Deep Learning: A subgroup of machine learning, this technology works by using algorithms and models like the network in the human brain. These networks are known as artificial neural networks which can process unstructured data to identify patterns.
- Natural Language Processing (NLP): This area of AI provides the machine with the ability to understand and process human language. It can read, understand, and synthesize information from human languages to produce valuable insights.
Through these AI technologies, pharmaceutical companies have already begun leveraging on each individual capability to make improvements to the drug discovery and development process. In hopes to mitigate the high cost and produce market ready new medicines efficiently and effectively.
Steps in Drug Discovery
The drug discovery process typically has four key stages and takes place over five to six years before the start of clinical development. These main steps however do not include the initial drug target identification and assay development steps which are also crucial and are mainly done in academia and research centres.
Stage 1: Screening and Target Identification
At this stage, the potential target of drug intervention is determined to identify any possible “druggable” points of the specific disease involved. Identifying potential therapeutic targets such as proteins, nucleic acids, or genes that are associated with the pathophysiology of the disease is a key part of this stage. For a therapeutic target to be viable for intervention, it should meet key clinical and commercial requirements, as well as demonstrate safety and efficacy in treatment of the disease.
Stage 2: Target Validation (Hit to Lead Selection)
Validation of the target (Hit) is important for showing how it plays a part in the development of the specified disease. This is done together with validation of the drug molecule’s (Lead) efficacy and toxicity based on testing with cell and animal models of the disease. This stage can be further divided into two steps; reproducibility and introducing variation to the drug-target environment.
Stage 3: Lead Identification
The lead or drug-like molecule is identified and tested in assays to determine its specificity, affinity and selectivity for the therapeutic target receptor of the specified disease. The structural and feasibility of the lead for the target are defined through in vivo assessment. This process is crucial in decreasing the risk of failure in the drug development process when the lead molecule is formulated into a drug. The lead’s pharmacokinetic profile and cytotoxicity are also tested at this stage.
Stage 4: Lead Optimization and Product Characterization
Following the identification of the lead molecule, it is then optimized to formulate a drug candidate which is a repetitive process to design the potential new medicine. Lead molecules are examined for its properties which include, selectivity and binding mechanisms. This final stage is used to determiner whether the lead is able to retain its key properties even when its structure is altered during the production of a pre-clinical drug candidate. Important parameters evaluated during this stage are pharmacodynamic, pharmacokinetic, and toxicological properties. Optimization of the lead would require data regarding the toxicity, efficacy, stability and bioavailability of the lead molecule.
Considering these four main stages of drug discovery, it is a long and gruelling process of identifying a molecule for further drug development. Not to mention the high cost would drive established pharmaceutical companies as well as biotech start-ups to look to AI-enabled tools for solutions in mitigating cost and shortening the timelines for drug discovery.
Technologies in AI Used in Drug Discovery
Many major biopharmaceutical companies have already begun researching on how AI technologies can be used to tackle the challenges in various parts of the drug discovery process. As of 2019, most of the key biopharmaceutical companies have already publicly declared deals for using AI technologies for drug discovery. (Figure 1)
In October 2017, the Accelerating Therapeutics for Opportunities in Medicine (ATOM) consortium was established through public-private partnerships with the aim of changing the drug discovery process by accelerating development of more effective therapies for patients. To achieve its goal of developing a validated drug discovery platform that optimizes pharmacokinetics, toxicity, protein-ligand interactions, systems-level models, molecular design, and novel compound generation, all at the same time, ATOM leveraged on machine learning methods. Launching its ATOM Modelling Pipeline (AMPL) it uses machine learning to develop models from a wide range of historical drug discovery data. This pipeline was created to be modular and developed to work with generative algorithm for optimizing multiple parameters required for drug discovery. Through its AI approach, ATOM is able to examine the pharmacology and therapeutic potential of the drug at the same time. This would in turn help to reduce the number of molecules selected for experimental validation, accelerating the process of drug discovery and subsequently drug development.
In a report by Signify Research, venture capital investment has reached US$ 5.2 billion between 2012 and 2019 for AI in drug development and clinical trials despite a dip in 2019. AI technologies have been explored for a variety of applications in drug discovery, this includes data mining to identify new drug candidates, identifying new disease targets, designing new drug candidates, and drug optimisation. Capabilities of AI algorithms has enabled time savings in sifting through large amounts of data, letting the computer do the work of filtering out any key information from relevant databases.
In 2016, Pfizer and IBM Watson Health collaborated to boost drug discovery for research in immuno-oncology and development of cancer treatment. The IBM Watson was launched as a cloud-based platform for drug discovery that provides machine learning, natural language processing technologies for the identification of new drug targets for research and selecting new strategies in immuno-oncology.
“Pfizer remains committed to staying at the forefront of immuno-oncology research,” said Mikael Dolsten, President, Pfizer Worldwide Research & Development.
“With the incredible volume of data and literature available in this complex field, we believe that tapping into advanced technologies can help our scientific experts more rapidly identify novel combinations of immune-modulating agents. We are hopeful that by leveraging Watson’s cognitive capabilities in our drug discovery efforts, we will be able to bring promising new immuno-oncology therapeutics to patients more quickly.”
IBM’s Watson for Drug Discovery is able to take in 25 million abstracts of scientific articles, more than one million full-text medical journal articles, and four million patents. It is continually updated and is also able to be optimized to process data from unstructured data such as lab reports to identify patterns to formulate valuable information.
Founded in 2012, Exscientia looks to automate the drug discovery process by using AI-driven tools. Since its for establishment, it has made great progress in collaborating and partnering with other pharmaceutical companies to provide AI solutions to accelerate the drug discovery process in a number of disease areas. Some recent collaborations include one with Bayer AG that started in early 2020. This collaboration looks to apply AI-driven drug discovery in the field of cardiovascular disease and oncology.
Gero, a Singapore-based start-up is making us of its AI-modelling platform for drug discovery against diseases related to ageing.
With all the potential of applying AI in boosting the drug discovery process to be more cost-effective and efficient, a concerted effort has to be made by research centres and biopharmaceutical companies. In a report by the Deloitte Centre for Health Solutions on AI technologies for drug discovery, the need for collaborations and consortiums was highlighted was one of the key considerations for the use of AI solutions. Partnerships between technology companies to provide these AI expertise and pharmaceutical companies would also be required for the paradigm shift to adopting AI-solutions in drug discovery. [APBN]
References
- Paul, S., Mytelka, D., Dunwiddie, C. et al. How to improve R&D productivity: the pharmaceutical industry’s grand challenge. Nat Rev Drug Discov 9, 203–214 (2010). https://doi.org/10.1038/nrd3078
- Deloitte Insights. Intelligent Drug Discovery Powered by AI. (n.d) Retrieved from: https://www2.deloitte.com/content/dam/insights/us/articles/32961_intelligent-drug-discovery/DI_Intelligent-Drug-Discovery.pdf
- Mohs, R. C., & Greig, N. H. (2017). Drug discovery and development: Role of basic biological research. Alzheimer’s & dementia (New York, N. Y.), 3(4), 651–657. https://doi.org/10.1016/j.trci.2017.10.005
- Deore, A., Dhumane, J., Wagh, R., & Sonawane, R. (2019). The Stages of Drug Discovery and Development Process. Asian Journal of Pharmaceutical Research and Development, 7, 62–67. https://doi.org/10.22270/ajprd.v7i6.616
- Mak, K.-K., & Pichika, M. R. (2019). Artificial intelligence in drug development: present status and future prospects. Drug Discovery Today, 24(3), 773–780. https://doi.org/https://doi.org/10.1016/j.drudis.2018.11.014
- Hinkson, I. V, Madej, B., & Stahlberg, E. A. (2020). Accelerating Therapeutics for Opportunities in Medicine: A Paradigm Shift in Drug Discovery In Frontiers in Pharmacology (Vol. 11, p. 770). https://www.frontiersin.org/article/10.3389/fphar.2020.00770
- Ulrik Kristensen, Signify Research, VC Funding for AI in Drug Development & Clinical Trials Hits $5.2B. (December 18, 2019). Retrieved from: https://www.signifyresearch.net/healthcare-it/vc-funding-ai-drug-development-clinical-trials-hits-5-2b/