Q/A: IQVIA’s Global Lead Talks Unlocking AI for Drug Repurposing

In an interview with HIT Consultant, Nathan Sommerford, Global Lead for Drug Discovery and Development, IQVIA talks how unlocking artificial intelligence (AI) in drug repurposing is transforming pharma and biotech companies.

What are some of the challenges and problems associated with drug discovery “from scratch”? 

Nathan Sommerford, Global Lead for Drug Discovery and Development, IQVIA: In the traditional drug discovery and development paradigm, it can cost up to $1.8 billion and take 10 years or more to develop a new medicine and successfully bring it to market, with drug discovery and screening stages alone taking up to four years. 

Often, traditional approaches are not driven by the strength of hypothesis but are led by the therapy area. Sponsors may unintentionally limit asset potential within the walls of the particular therapeutic space and related business goals. 

Given these factors and more, traditional approaches to drug discovery and development can be a difficult and long process for all stakeholders, especially patients waiting for life-changing treatments. 

Though there is a vast level of chemical space to foster the development of drug molecules, a lack of advanced technologies can limit the traditional drug development process, making it time consuming and expensive for pharmaceutical and biotech companies to pursue. And, unfortunately, as up to 90 percent of assets do not make it to market, traditional approaches are high risk for these companies as well.

How does repurposing compare (favorably and less so) to from-the-drawing-board discovery? And, how can use of artificial intelligence/machine learning and other advanced analytical tools help improve the repurposing process and empower drug development teams to achieve what isn’t possible (or as straightforward, perhaps) with other means?

Sommerford: In recent years, AI-based approaches to drug repurposing are growing and grabbing the attention of trial sponsors because these data-driven and tech-enabled strategies are reducing drug discovery time frames and failures rates. AI-based repurposing techniques are offering sponsors a speedier option to identify data-backed patterns for potential drug usage that simply may not be feasible through manual processing and analysis. With these tech-enabled approaches doing the heavy lifting, there is an opportunity for sponsors and the industry to scan the vast chemical space available with quickness and accuracy and then funnel assets with potential to the forefront of development.

By extracting insights and evidence from the massive breadth of datasets available to us through electronic medical records, claims data, scientific literature, genomics data and more, AI/ML techniques can help uncover hidden patterns that tap into initially unsuccessful molecules, providing another shot on goal for the asset against a different disease target and indication. For investigational drugs across phases of development that fail to meet endpoints, sponsors have the potential to course correct whether it is within a live trial or otherwise and ensure the investments to date are not fruitless. 

Additionally, these techniques are flexible enough to be used for approved drugs on market to optimize the asset against other indications or different therapy areas with the added benefit of already having the safety profile of approved drugs from previous trials. Repurposing a successful drug relatively quickly allows sponsors to open up their focus to patients in another therapy area who need viable options. 

However, there are limitations to consider regarding drug repurposing. For one, there has to be assurance that the AI/ML approach used is couched in the correct science and guided by medical and clinical experts when deciding how to develop a molecule for patients. Integrating data science expertise into the broader development team is key to ensure AI/ML techniques are applied appropriately within the fuller process, leveraging experience and knowledge to identify patterns and extract meaningful insights for successful repurposing activities.  

It is also important to keep in mind that with traditional drug development, knowledge of failed assets, including why the drug failed, is often limited and unpublished. And, in many cases, it is difficult to secure all necessary data to adequately analyze a drug’s clinical benefits and alternative uses due to study design, lack of endpoints or a small number of enrolled patients—making it key to ensure the approach used accounts for gaps in information.  

What are some of the varying systematic approaches to repurposing?

Sommerford: Generally, there are several categories of repurposing strategies, which sponsors can consider depending on where they are in the asset(s) lifecycle: 

For Phases 1-3: 

• When molecules did not make it through the initial development stage. 
  • When molecules currently in clinical trial stage are failing to meet endpoints. 

Post-market: 

• When using an approved drug for an indication extension or an adjacent indication in the same therapeutic area.
  • When an existing drug is reused for a different indication that what it is currently approved to address an unmet patient need. 

What about personalized therapies—how can harnessing AI/ML improve upon more conventional processes?

Sommerford: In having access to massive amounts of genomics data, the industry has an opportunity to gain meaningful insights, further personalizing medicine to a patient’s particular genomic makeup. However, genomics data is complex, and personalizing therapies requires accurately examining disease sub-types and the investigational compound’s sensitivity to various genomics profiles. As noted previously, this process may not be feasible to do manually, and when possible manually, it will be laborious, time-intensive and costly. This could potentially cause sponsors to forego leveraging genomics data and creating a limited view of asset capability. 

Using a ML-based approach, there is a chance to glean valuable insights of how genomics data interacts with data regarding compounds, proteins, EMRs, cellular-level images and more. These specific insights can help identify novel and notable patterns regarding the compounds potential capabilities such as drug response and more. 

What should professionals looking to dive into AI for drug repurposing and personalized medicine know?

Sommerford: AI/ML is positively impacting the drug development paradigm in pharmaceutical research and development. Enhancing these paradigm-shifting approaches requires moderate investment from sponsors for a substantial potential return on investment. Working in lockstep with clinicians and research scientists, AI/ML can provide a much-needed augmentation to how drugs are developed. Along with opening up clinical development programs for one asset, these applications can help reach patients faster by increasing the speed and success of development while reducing cost of molecule discovery to market entry. 

genomics
medical
biotech
pharma
artificial intelligence
ai
emrs
machine learning
personalized medicine
medicine
pharmaceutical