By: David Ostrowsky
Without question, artificial intelligence beholds many downright frightening risks when implemented in the healthcare field. Without proper oversight, AI could have the final say in depriving a patient of a necessary procedure or steering them towards a suboptimal physician. Or the still nascent technology, if unchecked, could account for a HIPAA breach with the slapdash handling of PHI. Yet, that being said, AI is undoubtedly here to stay and as far as the healthcare industry is concerned, this technological revolution is brimming with astonishing potential, as evidenced by the recent development of AI expediting the repurposing of existing drugs for treatment of rare diseases.
No the process of repurposing already FDA approved drugs for new conditions such as rare and aggressive cancers, fatal inflammatory disorders and complex neurological conditions is not new. However, the application of A/I fueled Machine Learning to speed up the process is nothing short of groundbreaking and could very well have a profound impact on the tens of millions of Americans who suffer from a rare disease, which is defined by The National Institutes of Health as one which affects fewer than 200,000 people in the United States.
When speaking to the New York Times earlier this spring, Donald C. Lo, the former head of therapeutic development at the National Center for Advancing Translational Sciences and a scientific lead at Remedi4All, a group focused on drug repurposing, explained that there is a “treasure trove of medicine that could be used for so many other diseases. We just didn’t have a systematic way of looking at it. It’s essentially almost silly not to try this, because these drugs are already approved. You can already buy them at the pharmacy.”
Though hundreds of millions around the globe may be grappling with a terrifying, truly life-jeopardizing “rare disease,” the inconvenient truth is that over 90 percent of rare diseases currently have no approved treatments, largely because titans of the pharmaceutical industry are reluctant to invest significant resources towards the discovery of a treatment that, in all likelihood, will only be utilized by a relatively miniscule portion of the population.
So if humans are unable to take full advantage of the potential of drug repurposing, the robots may be able to – more specifically, a splashy new AI model that goes by the name of TxGNN, the first one developed with the express purpose of identifying drugs for extremely rare diseases for which there are no known treatments. Thus far, TxGNN, overseen by scientists at Harvard Medical School, has identified drug candidates from current medicines for well over 17,000 diseases – a staggering total that stands as the largest number of diseases any one particular AI model has thus far been responsible for. To their credit, the Harvard researchers have made TxGNN available at no cost while encouraging clinicians and scientists to apply it in their pursuit of finding new therapies.
“With this tool we aim to identify new therapies across the disease spectrum but when it comes to rare, ultrarare, and neglected conditions, we foresee this model could help close, or at least narrow, a gap that creates serious health disparities,” noted lead researcher Marinka Zitnik, assistant professor of biomedical informatics in the Blavatnik Institute at Harvard Medical School, on the HMS official website back in September. “This is precisely where we see the promise of AI in reducing the global disease burden, in finding new uses for existing drugs, which is also a faster and more cost-effective way to develop therapies than designing new drugs from scratch.”
By pinpointing drug candidates from approximately 8,000 medicines (both FDA-approved medicines and experimental ones currently in clinical trials) for 17,080 diseases, TxGNN was close to 50 percent more accurate, on average, at identifying drug candidates when compared to other industry-leading AI models for drug repurposing. According to Harvard Medical School, TxGNN, which has two central features, one for identifying treatment candidates along with possible side effects and another one detailing the reasoning behind the decision, was also 35 percent more accurate in predicting what drugs would have contraindications (particular situations in which a medication, procedure, or surgery should be avoided because it could potentially harm the patient.)
The extent to which TxGNN and other similarly next generation AI models for drug repurposing has a transformative impact on the industry remains to be seen. On the surface, repurposing already approved drugs represents an intriguing way to generate new treatments because the technique relies on medicines that have already been studied, have well-comprehended safety profiles, and have been through the regulatory approval process. Viewed through a different lens, should this gargantuan development come to fruition, it will be interesting to see if there could be progress made in lessening social health disparities as countless more socioeconomically disadvantaged patients could have access to specialized drugs that would otherwise be financially prohibitive.