I wrote the following piece for my class last term, “Assessing Artificial Intelligence and Computational Technologies,” which was taught by Professor Daniel Linna. In class, we identified and assessed business opportunities and societal risks/benefits associated with AI. We were assigned to write a short essay styled as a blog post on a topic related to AI and computational technologies. I chose to write about AI and pharma because I hope to one day work in the pharmaceutical industry and I’m also very interested in how AI is helping bring new drugs to market.
Can Artificial Intelligence Help Bring Pharmaceuticals to The Masses?
Artificial Intelligence (AI) has empowered many industries. From accelerating automation on the assembly line to creating predictive analytics to promoting social justice, AI has transformed operational efficiency across a multitude of platforms. This is especially evident in industries where it can take upward of a decade for a product to be brought to market. Pharma is a good example; the pharmaceutical industry has received criticism for the time it takes to bring a therapeutic drug to the marketplace – on average, 8-12 years, with 3-6 of these years spent in pre-clinical drug discovery. This time outlay is matched by a multi-billion dollar financial investment. Add the time and money to the fact that companies in the pharma industry also eventually lose patent protection on many of their blockbuster drugs and it is easy to see that there is a need to innovate to stay competitive and profitable in this industry. Looking to innovate, many pharmaceutical companies have turned to deep learning. AI is a tool that the industry can leverage to increase accessibility to medicine. This adaptive technology can aid in pre-clinical discovery, thereby decreasing the time it takes for drugs to be brought to the marketplace. The use of deep learning algorithms in drug discovery can increase a drug manufacturer’s profit margins and can also increase global access to health.
The Long, Laborious Process of Drug Discovery
Drug discovery is the initial phase of R&D and involves the identification and optimization of potential new drugs. This process is slow, expensive, and often unsuccessful. Only about 1% of pre-clinical identifiable drug molecules and compounds are brought to the research phase for further testing in humans. Clinical testing is essential to assess drug safety and efficacy on the way to a drug receiving regulatory approval. Working through the drug discovery process involves a lot of risk for pharma companies. The costs associated with the search of a “hit” molecule that can treat the disease state and achieve the desired result upon retesting are tremendous. As many blockbuster drugs begin to lose their monopolies, big pharma companies are hungry for new innovations to help restock their pipelines and portfolios. Previously, it was much easier to develop a drug for a general health issue, but the move to more personalized medicine complicates the process. AI can help mitigate this issue and reduce initial costs, facilitate research developments, and increase the success rate by identifying promising drug candidates before entering the clinical stage.
Training These Deep Learning Models
By successfully creating AI models, drug firms can create predictive analytics regarding drug interactions on a molecular level. Compared to “normal” AI models, which are fed large amounts of data and recognize patterns, pharma companies are utilizing AI differently. Scientists now feed the algorithm a “sample problem (a molecule) and solutions (how the molecules will ultimately behave as a drug)” so that the software can develop its own computational approaches for producing solutions [1]. After the AI model has consistently been fed enough sample problems/solutions, it can begin to identify patterns that correlate with different interactions in the disease state. By identifying these specific protein and enzyme interactions, the software then pinpoints potential molecules that can bind to these proteins and stop the disease process. The software is predictive of how the same molecule will behave in the various system interactions within the human body. The AI software can develop and compile a “hit” list of drug candidates based upon predictive analytics. The result is to allow scientists to narrow their scope and concentrate time and resources on only the most promising candidates. We have begun to see some of the biggest players develop their own in-house AI software to aid in this, but many have turned to industry collaboration as well.
Out of a very competitive landscape, an industry alliance has formed to help derive biological insights from vast sets of data. The MELLODDY project (Machine Learning Ledger Orchestration for Drug Discovery), has created an AI platform of libraries of proprietary information. This allows each participating drug company the ability to comb through the research of other participating companies and use it to advance their own studies. Each player gets what they put into this project, while collaboratively building “models that are informed by the chemical and biological spaces of the combination of other partners” [2]. The MELLODDY project has yet to derive its full potential in aiding pharmaceutical firms, as they search for a molecule that has the potential to provide a therapeutic cure to a disease.
Future Implications
AI is not a shortcut for the entire R&D process, but can be used to improve the accuracy, predictability, and speed of drug discovery. It is an amplifier of innovation that will be able to battle many current and future industry counter-forces, while ultimately benefiting patients. As we begin to see the re-emergence of communicable diseases and pandemic risks, AI can help scientists discover a cure in a timely manner. Changing demographics, such as an aging population, along with the process of urbanization present challenges in the healthcare space. We must find ways to treat the chronic illnesses associated with aging and also some of the new disease trends resulting from clustering of the population in urban environments. AI can shift the social culture components of healthcare, such as access and attitudes. The current approach to drug pricing requires drugs to have a strong value proposition and pay for themselves. The use of AI can mitigate the initial investment of drug discovery, thereby impacting future retail drug pricing. An ancillary benefit would be the access of medicine available to the general public. In short, increased utilization of AI in drug development would result in a global good.
[1] https://www.nature.com/articles/d41586-019-03846-0
[2] https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(19)31401-1/fulltext
