By Anjali Samra
Artificial intelligence (AI) related tools are increasingly being used in the early drug discovery pipeline (Reda, Kaufmann & Delahaye-Duriez, 2020). Whilst some are looking towards the vast opportunities that are offered by such tools, there are also many that are opposed, waiting for a clear impact to be seen within the field. Although there have been many breakthroughs, this has also brought along a lot of new challenges (Schneider et al, 2019). This review will cover disruptive technological advances that have dominated discussion in the biotechnology and pharmaceutical industries today.
Currently it takes on average at least 10 years to bring a drug to market, with clinical trials alone taking around seven years. Moreover, the average cost of research and development for a successful drug is estimated to be $2.6 billion (DiMasi, Grabowski & Hansen, 2016). This high cost is related to the time-consuming process of introducing a drug to the marketplace alongside the high proportion of failures present in clinical trials (Smietana, Siatkowski & Moller, 2016). There is a growing need for more innovative approaches to quicken this very task and bring effective drugs at a lower cost for patients. Current challenges include multi-objective optimization problems associated with finding the optimal compound based on its potency, selectivity and potential off-target effects; datasets are also becoming larger and more complex, making it harder to analyse and make informed choices regarding selection (Baig et al, 2016). Therefore, there is much potential in AI as it could increase the accuracy of predictions on drug safety and efficacy as well as saving billions of dollars (Basile, Yahi & Tatonetti, 2019).
AI assisted small molecule drug design has been further fuelled by technological data processing opportunities. Up until now, high throughput screening (HTS) has an estimated hit rate of 0.01%, where a hit is a chemical compound that binds to the target and shows desired activity during initial screening (Zhu et al, 2013). An ambitious goal for this area is for de novo compound generation of the desired characteristics through AI methods, to cut down costs surrounding a full deck HTS (Popova, Isayev & Tropsha, 2018).
AI has already been applied vastly in areas like image recognition within MRI scans to detect subtle abnormalities that may otherwise be missed (Hainc et al, 2017). However, advances in drug discovery using this type of technology are much rarer; successful application relies on a drug making it through the rigor of a clinical trial and onto market. Therefore, decisions surrounding compound selection and clinical trial design must take precedence over cost or speed (Bender & Cortes-Ciriano, 2020). The biological aspect also needs to be taken into consideration through analysis of data related to safety and efficacy end points; without this compounds are more likely to fail in the trial stage due to a lack of in-vivo study as well as understanding of the underlying biology (Seyan, 2019).
Collaborations between biotechnology companies and big pharmaceutical organisations are needed to fully leverage the platforms developed by such drug discovery ventures. For example, a novel target for chronic kidney disease was developed through the partnership between BenevolentAI and AstraZeneca (BenevolentAI, 2021). Moreover, the first AI designed drug to enter human clinical trials in history by Exscientia alongside Sumitomo Dainippon Pharma marked a crucial milestone. The new compound was developed as treatment for obsessive compulsive disorder in just 1 year. The algorithms were able to design millions of molecules and identify which one would be the optimal candidate to test (Wakefield, 2020).
In the past few years there has been a striking increase in the number of companies centered around AI in drug discovery; this is evident by the rise in market going from $200 million in 2016 to over $700 million in 2018, with projections within the next five years reaching $20 billion (Deloitte, 2019). Moreover, there have been more investments from big technology companies, such as Microsoft as mentioned before. Also, Google DeepMind’s AI technology produced a novel disruptive deep learning algorithm that can use primary protein sequences to predict their 3D structure, surpassing current methods as well as experienced members within the field (Senior et al, 2020). A collaboration between the UK Vodafone Foundation and Imperial College London aimed to repurpose drugs using smartphones; the AI powered app would fuel cancer research by transferring unused processing power of the smartphones to a supercomputer. As a result, work that would originally have taken one computer 300 years was cut down to only 3 months producing 110 anti-cancer molecules (Veselkov et al, 2019).
Recently, Microsoft partnered with Novartis to accelerate drug development using AI. The combined expertise of both firms may provide an avenue to treatments for many diseases that are currently classified as uncurable. The result of increasing efficacy of data science methods through Microsoft in parallel with Novartis’ pharmaceutical research knowledge will help “answer some of the world’s most pressing health challenges”, as stated by Peter Lee , Corporate Vice President of Microsoft Healthcare (Yates-Roberts, 2019). One of ventures includes using Microsoft’s neural networks to create and screen promising molecules as drug candidates. This could pave the path for reinventing medicine as we know it, bringing affordability as well as accessibility to innovative healthcare (Lee, 2019).
There are many merits to AI but also challenges associated with its adoption. However, any hurdles can be overcome with enough research and time. AI has the ability to restructure medicine as we know it, to become more precise, predictive and preventative (Schork, 2019). Furthermore, treatments will soon become highly specific to their targets with consideration to individual genetic backgrounds and toxicity profiles (Nair, 2010). There are boundless opportunities within the field to improve drug discovery platforms and bring medicine to patients quicker as well as meet unmet needs within certain disease categories that do not currently have effective therapies (Brasil et al, 2019). Thus, to unlock the full capabilities of AI in drug discovery, the issues associated with the type of data used must be addressed (Bender & Cortes-Ciriano, 2020).
References:
Reda, C., Kaufmann, E. & Delahaye-Duriez, A. (2020). Machine learning applications in drug development. Computations and Structural Biotechnology Journal. 18, 241-252. Available from: doi.org/10.1016/j.csbj.2019.12.006
Schneider, P., Walters, W., Plowright, A., Sieroka, N., Listgarten, J., Goodnow, R., Fisher, J., Jansen, J., Duca, J., Rush, T., Zentgraf, M., Hill, J., Krutoholow, E., Kohler, M., Blaney, J., Funatsu, K., Luebkemann, C. & Schneider, G. (2019). Rethinking drug design in the artificial intelligence era. Nature Reviews Drug Discovery. 19(5), 353-364. Available from: doi.org/10.1038/s41573-019-0050-3
DiMasi J. A., Grabowski H. G & Hansen R. A. (2016) Innovation in the pharmaceutical industry: new estimates of R&D costs. Journal of Health Economics. 47, 20-33. Available from: doi: 10.1016/j.jhealeco.2016.01.012
Smietana, K., Siatkowski, M. & Møller, M. (2016). Trends in clinical success rates. Nature Reviews Drug Discovery. 15(6), 379-380. Available from: doi: 10.1038/nrd.2016.85
Baig, M. H., Ahmad, K., Roy, S., Ashraf, J. M., Adil, M., Siddiqui, M. H., Khan, S., Kamal, M. A., Provaznik, I. & Choi, I. (2016). Computer Aided Drug Design: Success and Limitations. Current Pharmaceutical Designs. 22, 572-581. Available from: doi: 10.2174/1381612822666151125000550
Basile, A. O., Yahi, A. & Tatonetti N. P. (2019). Artificial Intelligence for Drug Toxicity and Safety. Trends in Pharmacological Sciences. 40(9), 624-635. Available from: doi.org/10.1016/j.tips.2019.07.005
Zhu, T., Cao, S., Su, P., Patel, R., Shah, D., Chokshi, H., Szukala, R., Johnson, M. & Hevener, K. (2013). Hit Identification and Optimization in Virtual Screening: Practical Recommendations Based on a Critical Literature Analysis. Journal of Medicinal Chemistry. 56(17), 6560-6572. Available from: doi: 10.1021/jm301916b
Popova, M., Isayev, O. & Tropsha, A. (2018). Deep reinforcement learning for de novo drug design. Science Advances, 4(7), 7885. Available from: doi: 10.1126/sciadv.aap7885
Hainc, N., Federau, C., Stieltjes, B., Blatow, M., Bink, A. & Stippich, C (2017). The Bright, Artificial Intelligence-Augmented Future of Neuroimaging Reading. Frontiers in Neurology. 8, 489. Available from: doi: 10.3389/fneur.2017.00489
Bender, A. & Cortés-Ciriano, I. (2020). Artificial intelligence in drug discovery: what is realistic, what are illusions? Part 1: Ways to make an impact, and why we are not there yet. Drug Discovery Today. S1359-6446(20), 30527-4. Available from: doi.org/10.1016/j.drudis.2020.12.009
Seyhan, A. (2019). Lost in translation: the valley of death across preclinical and clinical divide – identification of problems and overcoming obstacles. Translational Medicine Communications. 4(18). Available from: doi.org/10.1186/s41231-019-0050-7
BenevolentAI. (2021). BenevolentAI and AstraZeneca achieve collaboration milestone with novel AI-generated chronic kidney disease target. Available from: https://www.benevolent.com/news/benevolentai-and-astrazeneca-achieve-collaboration-milestone-with-novel-ai-generated-chronic-kidney-disease-target [Accessed: 15th March 2021].
Wakefield, A. (2020). Artificial intelligence-created medicine to be used on humans for the first time. Available from: https://www.bbc.co.uk/news/technology-51315462. [Accessed 15th March 2021].
Deloitte. (2019). Intelligent drug discovery powered by AI. Available from: https://www2.deloitte.com/content/dam/insights/us/articles/32961_intelligent-drug-discovery/DI_Intelligent-Drug-Discovery.pdf [Accessed: 15th March 2021].
Senior, A., Evans, R., Jumper, J., Kirkpatrick, J., Sifre, L., Green, T., Qin, C., Žídek, A., Nelson, A., Bridgland, A., Penedones, H., Petersen, S., Simonyan, K., Crossan, S., Kohli, P., Jones, D., Silver, D., Kavukcuoglu, K. & Hassabis, D. (2020). Improved protein structure prediction using potentials from deep learning. Nature. 577(7792), 706-710. Available from: doi.org/10.1038/s41586-019-1923-7
Veselkov, K., Gonzalez, G., Aljifri, S., Galea, D., Mirnezami, R., Youssef, J., Bronstein, M. & Laponogov, I., 2019. HyperFoods: Machine intelligent mapping of cancer-beating molecules in foods. Scientific Reports. 9(1). Available from: doi.org/10.1038/s41598-019-45349-y
Yates-Robertson, E. (2019). Microsoft and Novartis partner to accelerate drug development. Available from: https://www.technologyrecord.com/Article/microsoft-and-novartis-partner-to-accelerate-drug-development-87323#:~:text=Microsoft%20is%20partnering%20with%20pharmaceutical,artificial%20intelligence%20(AI)%20technology.&text=A%20central%20part%20of%20the,the%20vast%20amounts%20of%20data. [Accessed: 16th March 2021]
Lee, P. (2019). Bringing together deep bioscience and AI to help patients worldwide: Novartis and Microsoft work to reinvent treatment discovery and development. Available from: https://blogs.microsoft.com/blog/2019/10/01/bringing-together-deep-bioscience-and-ai-to-help-patients-worldwide-novartis-and-microsoft-work-to-reinvent-treatment-discovery-and-development/ [Accessed: 16th March 2021]
Schork, N.J. (2019). Artificial intelligence and personalised medicine. Cancer Treat Res. 178, 265-283. Available from: doi: 10.1007/978-3-030-16391-4_11
Nair, S., 2010. Personalized medicine: Striding from genes to medicines. Perspectives in Clinical Research. 1(4), 146. Available: doi: 10.4103/2229-3485.71775
Brasil, S., Pascoal, C., Francisco, R., dos Reis Ferreira, V., A. Videira, P. and Valadão, G., 2019. Artificial Intelligence (AI) in Rare Diseases: Is the Future Brighter?. Genes. 10(12), 978. Available from: doi: 10.3390/genes10120978
Imperial Bioscience Review 