Drug Repurposing: Expediting Discovery Through a Novel Algorithmic Framework
As the COVID-19 pandemic continues to stretch the ability and capacity of health care systems—from curbing infection rates, treating severe illness, understanding new variants, to distributing and administering limited vaccine doses—gaps and vulnerabilities within those same systems continue to emerge and slow the response to a highly contagious, infectious disease, prompting health care experts to find creative ways to forge ahead.
One vulnerability of our health care system is the traditionally slow process of developing new therapeutics and vaccines to treat and prevent emerging diseases such as COVID-19, and the need for new technologies to allow for a more rapid response. Fortunately, new technologies have been waiting in the wings. New mRNA-based vaccine technology has shown the power of advances in vaccine development that, along with political will and monies, has produced highly-effective and safe vaccines in record time. On the therapeutic front, repurposing approved drugs to treat symptoms of COVID-19 is under ongoing investigation as an essential way to offer a quicker and more cost-effective ways to get safe and efficacious drugs to people who need them now, compared with looking for new agents through traditional drug discovery methods.
Drug repurposing has gained traction over the years, with data showing that repurposing comprises about one-third of drug approvals in recent years and generates about 25% of the yearly revenue for pharmaceutical companies.
As discussed in an editorial by Talevi and Bellera,1 many of the most well-known drugs that have been repurposed over the years, such as sildenafil, aspirin, and valproic acid, were identified through serendipitous observation of their efficacy for conditions other than what they were approved for. Newer, more organized methods are now being employed that use advances in technology to expedite drug repurposing, with a growing number of databases of compounds identified or under investigation (see sidebar of databases on drug repurposing).
Among these technologies are high-powered computation and artificial intelligence (AI) tools such as machine and deep learning to process large amounts of data. Using these tools, investigators from The Ohio State University recently reported results on their ability to identify repurposed drugs effective for cardiovascular disease based on a computational framework they developed using the powers of AI to process real-world data to identify drugs for repurposing. Importantly, the framework can be adapted to other diseases and disease pathways, such as COVID-19.
Using Real-World Data With AI to Identify Drugs for Repurposing
Published in Nature Machine Intelligence,2 the study describes an algorithmic framework developed by Dr Zhang and colleagues for generating and testing multiple candidates for drug repurposing. The framework is based on a retrospective analysis of real-world data using AI tools that emulate data obtained from randomized clinical trials.
“We are the first team to introduce the use of the deep learning algorithm to handle the real-world data, emulate clinical trials, and identify new uses for existing medications,” said senior author of the study Ping Zhang, PhD, assistant professor, department of computer science and engineering, and department of biomedical informatics, The Ohio State University, Columbus, OH.
The benefit of using real-world data over traditionally used preclinical data for drug discovery, he said, is that existing real-world data, such as electronic health records, insurance claims, and billing activities, requires lower costs, can scale to a larger number of patients, and better represents the heterogeneity in the population.
In addition, he said that preclinical data often fail to correlate well with the therapeutic efficacy in drug development with studies showing that only 30% of all compounds found effective in cell assays would work in animals and only 5% in humans.
“Real-world data are direct observations from clinical patients that may be seen as valuable readouts of drug effects directly on human bodies,” said Dr Zhang.
One big drawback, however, is the many types of bias in real-world data such as selection bias and indication bias. To overcome that drawback, Dr Zhang and colleagues used tools of AI—causal inference algorithms with deep learning methods—to process large amounts of patient information that can affect how drugs work in the body (ie, confounders such as sex, age, race, disease severity, comorbidities) and control for these multiple confounds in a way that emulates clinical trials.
To test their framework, the investigators used it for identifying drug repurposing for a cohort of 1.2 million patients with coronary artery disease (CAD). Real-world data on the cohort was obtained from a large-scale medical claims database and was inputted into the algorithmic framework to which AI was applied to process the information.
The study showed that the framework successfully retrieved three of four known drugs (ie, amlodipine, diltiazem, and rosuvastatin) for CAD, according to Dr Zhang. In addition, the framework also identified six additional drugs that are not currently indicated for CAD, but were shown to improve disease outcomes (metroprolol, fenofibrate, hydrochlorothiazide, pravastatin, simvastatin, and valstartan), and therefore potential drug repurposing candidates.
Other findings showed that some drugs, such as metformin and escitalopram not previously associated with CAD, showed therapeutic effect, said Dr Zhang. In addition, some drugs not found to be effective as monotherapy for CAD (ie, lisinopril and atorvastatin) were found to substantially improve CAD when combined.
Dr Zhang said that these findings pave the way for using real-world data and AI for drug repurposing but remain at the “stage of hypothesis generation” given the mandate by the Food and Drug Administration for evidence from clinical trials for drug approval.
Dr Zhang underscored that the framework can be applied to any disease if the cohort and outcome are defined, and said he’d love to collaborate with domain experts to develop drugs for unmet needs such as for rare diseases without any treatment and COVID-19. If applied to COVID-19, for example, he said the cohort would be all COVID-19 patients and the outcomes could be those with serious disease such as acute respiratory distress syndrome and mortality. “Our method can reduce confounders, emulate clinical trials for candidate drugs, and identify potential drugs for reducing the severity of COVID-19.”
Challenges and Opportunities Ahead
Vivak Gupta, PhD, assistant professor, department of pharmaceutical sciences, College of Pharmacy & Health Sciences, St. John’s University, Queens, NY, who recently wrote on drug repurposing3 as a promising tool to accelerate the drug discovery process, believes drug repurposing is the future of drug discovery, and emphasized the importance of AI tools, given its ability to screen large amounts of data and its processing speed.
“AI models will provide a very efficient way to unscramble and scientifically organize the vast body of scientific data being generated and made public every day,” emphasized Dr Gupta, however, the importance of continually training computers in processing big data.
Other challenges, according to Mithun Rudrapal, PhD, associate professor, department of pharmaceutical chemistry, and Rasiklal M Dhariwal Institute of Pharmaceutical Education and Research, Chinchwad, Pune, India, include issues with data (eg, quality and quantity, labeling, accessibility, and security), people (eg, lack of understanding among nontechnical employees, lack of skill and talent, scarcity of field specialists), business (eg, lack of business alignment, difficulty in assessing vendors, rigid business models), infrastructure (ie, requirements for testing and experimentation), as well as issues of affordability and implementation.
Dr Rudrapal, who recently published a chapter on drug repurposing on the open-access site IntechOpen,4 suggested that meeting these challenges is important given the notable advantages of using computational approaches with AI and machine learning for drug repurposing. These advantages include the highly efficient ability of AI to process data, the lower risk of failure when using AI tools, and the less time needed for drug discovery and development.
Key opportunities for drug repurposing, according to Dr Rudrapal, are identifying drug molecules or therapeutic agents for particularly difficult-to-treat diseases, as well as rare and neglected diseases like parasitic diseases, cancer, and infectious diseases like COVID-19.
“Drug repurposing also has ample opportunities in some other therapeutic and pharmacological areas such as personalized medicines and precision medicine, system medicine, and network pharmacology,” he said.
Given the many benefits and advantages of these approaches over older methods, Dr Rudrapal said that, “It is extremely important for managed care people such as health care providers, decision-makers, and pharmacists to understand or get acquainted with newer computational tools and techniques including AI and machine learning for drug discovery and/or drug repurposing.”
Dr Gupta also emphasized that understanding drug repurposing is important for providers and other health care professionals so they can pass on that knowledge to their patients. “Being informed about drug repurposing enables providers to be aware of what other options may be available to patients with chronic diseases for whom treatments are lacking,” he said. “There are about 7000 rare diseases and millions of people who suffer from one of them and many have zero drugs, so if a provider has another option [like a repurposed drug] to offer I think they would be very excited because these drugs are already known to be safe.”
He also emphasized the role pharmacists play in drug repurposing, such as informing patients and providers about potential drug interactions when introducing a repurposed drug into the care of a patient taking other medications.
References:
- Talevi A, Bellera CL. Challenges and opportunities with drug repurposing: finding strategies to find alternative uses of therapeutics, Expert Opinion on Drug Discovery. 2020;15(4):397-401. doi:10.1080/17460441.2020.1704729
- Liu R, Wei L, Zhang P. A deep learning framework for drug repurposing via emulating clinical trials on real-world patient data. Nature Machine Intelligence. 2021;3:68-75. doi:org/10.1038/s42256-020-00276-w
- Parvathaneni V, Kulkarni NS, Muth A, Gupta V. Drug repurposing: a promising tool to accelerate the drug discovery process. Drug Discovery Today. 2019;24(10):2076-2085. https://doi.org/10.1016/j.drudis.2019.06.014
- Rudrapal M, Khairnar SJ, Jadhave AG. Drug Repurposing (DR): An emerging approach in drug discovery [published online July 13, 2020]. Drug Repurposing - Hypothesis, Molecular Aspects and Therapeutic Applications. doi:10.5772/intechopen.93193
