By Clarice Tse
The wide application of artificial intelligence in various industries and economic sectors has been under the limelight in recent years. The integration of AI, such as machine learning and deep learning, in traditional industrial sectors have enhanced the efficiency and productivity of these industries. The pharmaceutical industry is no exception to the enjoyment of the boons of AI integration. Many experts have admitted the promising future of AI in drug discovery and development, why so? In this article, we will look into the impactful use of AI in diverse areas of the pharmaceutical sector.
The pharmaceutical industry began in the 19th century and bloomed in the 20th century as there was a great leap in the development of modern medicine. As time went on, more restrictions and guidelines were established to ensure drug safety and quality. Moreover, as more drugs are developed, it has become harder to develop new drugs for particular diseases that are better than the existing drugs available. Easily tractable drug targets are already targeted in existing drugs, leaving researchers with more challenging drug targets, which takes more time to research on. The emerging phenomenon of inefficiency in research and development of drugs is evident from the declining rate in the number of drugs approved per billion $US spent on R&D. The cost of developing and discovering a drug was US$800 million in 2001 and skyrocketed to around US$3 billion. Therefore, there is strong demand for AI to address the current challenges in the pharmaceutical industry (Mak and Pichika, 2019).
The use of AI in drug discovery and development revolves around the fundamental paradigms of machine learning and deep learning. Machine learning uses statistical methods with the ability to learn with or without being explicitly programmed. In supervised machine learning, data input and results output are used for machine learning to build a predictive model which can be used in disease diagnosis and prediction of drug efficacy and ADMET (absorption, distribution, metabolism, excretion and toxicity). On the other hand, unsupervised machine learning, which only involves grouping and interpreting data based solely on input data, allows the discovery of new disease subtypes from disease target discoveries. This technology is ubiquitously applied in different stages of drug development such as identification and validation of drug targets, drug design, drug repurposing, decision making processes in clinical trials and evaluation of biomedicine information (Mak and Pichika, 2019).
For example, in the first stage of drug development, AI has been harnessed to develop new lead compounds in silico that demonstrate desired activity. Machine learning techniques contribute to the identification of target-specific molecules and the association of the molecule with the target while optimizing its safety and efficacy. Through induction and deduction of available data obtained from fragment screening, computational modelling of compounds, AI can optimize hit and lead compounds. Eventually, de novo designs of compounds are synthesized in silico and screening models are virtual, which will replace the biochemical tests for efficacy and toxicity of the drug (Lake, 2019). Thanks to AI, the complexity of wet lab experiments that evaluate drug efficacy can be taken over by the convenience of virtual screening on the computer, which is not only more efficient, but also more cost-effective.
Another subfield of machine learning, called deep learning, uses artificial neural networks which is a biologically inspired programming paradigm that adapts experimental data and learns from them. These comprise a set of sophisticated computing elements ‘perceptions’ analogous to biological neurons that are interconnected; they imitate the transmission of electrical impulses in the human brain. This enables us to discover new compounds that can potentially become new drugs and uncover drugs that can be more potent when combined with other drugs for personalized medicine based on genetic markers ((Bajorath et al., 2020)
,Paul et al., 2020). For instance, Insilico Medicine- a company that provides AI solutions to top pharma and biotechnology companies developed used deep learning to design small molecule drugs for a wide range of human diseases, such as cancer, fibrosis and immunological diseases. Most recently, the company utilized large datasets to train its deep learning computational model and performed in silico design of de novo molecular structures which are novel potential chemotypes targeting SARS-CoV2 main protease (Mpro) (Insilico & Nanome Use Virtual Reality to Refine New AI-Generated Drugs for COVID-19, 2020).
Other than evaluating the quality of drugs, involving in drug designs and discovering potential hits and leads, AI also contributes to drug synthesis and the manufacturing sector. During drug synthesis, scientists would have to first analyse the target compounds and break them down into building blocks that can be easily purchased and prepared. Then, they interrogate the vast number of organic reactions and identify those that would convert the building blocks back to the target compound. This process is brain-wrecking, so using AI to tackle these processes would be much faster and more precise. Currently, there are only a few computer-aided organic compound synthesis systems that combine different deep neural networks to carry out these processes (Mak and Pichika, 2019). However, more development in this area of AI application is needed as it still lacks in some areas.
Overall, in the long term, AI poses promising effects in the pharmaceutical and biotechnological industry. The collaboration between the AI sector and the biological research sector is becoming more and more prevalent. In addition to that, the use of robotics also improved the productivity and workflow. However, there are no known, developed drugs that have utilized AI approaches. The reliability of artificial intelligence is also questioned by some experts as AI would have to prove themselves to be accurate over and over again in every decision they make or results they produce. Therefore, even though some may worry that AI would cause job losses, scientists and experts are still in demand.
References:
Mak, K. and Pichika, M., 2019. Artificial intelligence in drug development: present status and future prospects. Drug Discovery Today, 24(3), pp.773-780.
Lake, F., 2019. Artificial intelligence in drug discovery: what is new, and what is next?. Future Drug Discovery, 1(2), p.FDD19.
Bajorath, J., Kearnes, S., Walters, W., Meanwell, N., Georg, G. and Wang, S., 2020. Artificial Intelligence in Drug Discovery: Into the Great Wide Open. Journal of Medicinal Chemistry, 63(16), pp.8651-8652.
Paul, D., Sanap, G., Shenoy, S., Kalyane, D., Kalia, K. and Tekade, R., 2020. Artificial intelligence in drug discovery and development. Drug Discovery Today,.
Insilico.com. 2020. Insilico & Nanome Use Virtual Reality To Refine New AI-Generated Drugs For COVID-19. [online] Available at: <https://insilico.com/insiliconanome> [Accessed 10 November 2020].
Imperial Bioscience Review 