AI in drug development
Without a doubt, artificial intelligence could be very helpful in many areas of the healthcare industry. One such example is Lark, the personal health coach chatbot app. Julia Hu, CEO and co-founder of Lark Technologies, gave us an overview of the app, and its impact on people’s lives is evident. One can quickly see that having an app that can monitor the information your smartphone records and notify you if something is wrong is a great tool for disease prevention or chronic disease monitoring.
In addition to helping in prevention and monitoring of illnesses, AI can also help in areas of healthcare that may not seem as evident to the general public, but that are relevant for people nonetheless. Such is the case of the use of AI in the drug discovery process.
To understand how AI can optimize the procedure to develop a new drug, first we need to understand how drug discovery has been done until now. The first step in drug discovery consists in identifying the treatment targets for each disease, such as proteins that cause inflammation or help tumors grow. After the treatment targets have been identified, researchers examine thousands of compounds, both natural and synthetic, to try to find which of them interact strongly with the targets. This process of determining candidate medicines is a trial-and-error method that can take up to six years. The next step would be to file for clinical trials, which consist in a series of tests made to determine medicine safety, efficacy, and dosing. This step takes around six or seven years to complete. Finally, the drug must be filed for approval and only then will it be manufactured and available to people. From start to finish, this process can take around twelve to fifteen years and between 1.3 and 1.6 billion dollars just for a single drug [1].
This procedure seems quite inefficient, and the question of how can we speed it up arises. A possible answer is by implementing AI in the first step of the drug discovery process. One company that takes said approach is AtomWise, which developed a deep learning system called AtomNet [2]. This system is capable of analyzing and predicting how molecules will react with the treatment target, and propose potential medicine candidates. It does this much faster than conventional methods, cutting times from years down to months. AtomNet has already been put to the test, and it has been able to propose potential medicines for diseases like ebola and multiple sclerosis [3].
Another company that employs AI in the drug development process is BERG. This company takes a “back to biology approach” [4]. What this means is that the company focuses on improving the understanding of the biology of diseases by using AI systems. To do this, researchers gather biological data from healthy and unhealthy tissue samples, and then they analyze the information with an AI platform to establish the differences between healthy and disease environments. This approach makes it easier to determine the treatment target for each disease, one of the fundamental steps in drug discovery. BERG has had success by following this methodology; they currently have a drug candidate for an aggressive type of brain tumors undergoing clinical trials [5].
A faster and more accurate compound research process is not the only benefit from using AI in drug development. As mentioned earlier, the whole development process can cost around 1.3 and 1.6 billion dollars. It is estimated that about fifteen to twenty percent of the cost of a new drug corresponds to the compound research phase [6]. Using AI during this stage could significantly reduce costs, as expensive physical compound testing could be replaced by molecule interaction simulations.
As time goes by and AI systems become more sophisticated, we can expect for the drug discovery process to become faster and cheaper, leading to better treatments for well-known diseases as well as new treatments for complex illnesses that are currently incurable.
References
[1] http://www.phrma.org/video/drug-discovery
[2] http://www.atomwise.com/introducing-atomnet/
[3] https://singularityhub.com/2017/05/07/drug-discovery-ai-can-do-in-a-day-what-currently-takes-months/
[4] https://berghealth.com/company/
[6] https://www.wired.com/2017/03/supercomputers-stocking-next-generation-drug-pipelines/
2 comments on “AI in drug development”
Comments are closed.
Thank you for an interesting post! You mentioned that a couple of companies are employing AI in drug development and I would like to add that there is also a lot of academic research going on the same topic. It is indeed interesting how accurately machine learning can predict new drug-target interactions. In addition, one interesting direction in this area is the discovery of repurposing opportunities for existing drug compounds, i.e. predicting new targets for drugs that are already available on the market. In that way, the costs of laboratory testing and clinical trials can be avoided, thus making the drug development process even more efficient. There are a lot of interesting articles available on drug repurposing, such as:
1) Strittmatter S M. Overcoming Drug Development Bottlenecks With Repurposing: Old drugs learn new tricks. Nature Medicine 2014;20:590-591. (Available: http://www.nature.com/nm/journal/v20/n6/full/nm.3595.html?foxtrotcallback=true)
2) Sanseau P, Koehler J. Computational methods for drug repurposing. Briefings in Bioinformatics 2014;12(4):301-302. (Available: https://academic.oup.com/bib/article/12/4/301/241437/Editorial-Computational-methods-for-drug)
You mention an interesting point. Repurposing drugs could help reduce trial costs, as said drugs have already been determined safe for human use.
I think that aligns with what BERG is doing. By analyzing the biology of diseases with AI, there could be a better understanding of how they work, and with that information it may be possible to find an existing drug that could mitigate diseases other than its original target.
http://www.wired.co.uk/article/ai-cancer-drugs-berg-pharma-startup
Also, Atomwise, the company behind AtomNet, has also looked into existing drug repurposing to treat other diseases. They mention in their website that they are looking at drug repurposing to treat malaria.
http://www.atomwise.com/malaria/