Dr. Jo Varshney, founder and CEO of VeriSIM Life.
getty
Experts said it could not be done.
Traveling at the speed of sound (Mach 1.00) was impossible.
Scholars maintained that the pressures exerted on pilots and aircraft were too much to overcome. Yet on October 14, 1947, Chuck Yeager proved the experts wrong and became the fastest man alive, piercing an invisible brick wall in the sky and paving the way toward human space flight.
The sound barrier was commonly seen as unsurpassable for crewed aircraft, even though other objects such as the tip of a bullwhip and 19th-century firearms created supersonic velocity. Such is human nature; almost everything we do in modern life was considered impossible at one time or another.
In comparison, some people believe artificial intelligence (AI) and machine learning technologies cannot solve the complexities of biology and disease—despite evidence of how AI solves such complex tasks in our daily life, such as navigation, autonomous flight, language translation and more.
Why couldn’t AI help solve many of the seemingly incalculable challenges we face in drug discovery, where only 5 in 5,000 drugs that enter preclinical testing will progress to human testing?
We Can Do Better
I believe the answer is a resounding "it can." Today, thanks to the evolution of AI, the translational challenges in drug discovery are experiencing a Chuck Yeager moment.
A diverse range of AI applications is available to help develop new drugs. Improvements in technology are making the process more cost-effective, and it can potentially reduce the time a new drug reaches the patient significantly. Applying AI in the drug discovery process is not new. Still, its effectiveness in increasing the speed of finding new potential treatments and creating a pipeline of drugs to address more diseases is growing in unimaginable ways.
It has been said that coming up with new drugs is more complicated than rocket science because biology is so complex (sorry, Mr. Yeager). On average, it takes more than 10 years for just one medicine to make its way through the entire R&D process at an average cost of $2.6 billion. Ten years ago, every dollar that pharmaceutical companies invested in R&D saw a return of 10 cents; today, the yield is around three cents.
This is significant because it’s paramount to address the translational gap, better known as the “Valley of Death,” seen in drug development. The National Institutes of Health (NIH) reports that for every drug that receives FDA approval, more than 1,000 that were developed failed. Almost 50% of all experimental drugs fail in Phase III trials.
The good news is novel hybrid techniques combining AI with sophisticated quantitative modeling are bridging the translational gap we see in pharmacological sciences, biology and chemistry. This approach is helping pharmaceutical companies better de-risk and prioritize the most informative experiments they are conducting much earlier in the process. Much like mapping apps use AI to quickly deduce the best route between point A and point B (they do not simply “choose” from some pre-defined catalog of all possible routes—an impossible dataset to create), AI techniques can compensate for the lack of data that confounds standard modeling and prevents successful translation from the pre-clinical to the clinical stage. And with such a generalized and scalable learning approach, we can create a disease diagnostic system that enables more precise therapies to address the differences between patients.
Thank You, AI
With this machine learning-driven approach, pharmaceutical companies are now starting to preflight advancements in medicine with far fewer redundant trials and less animal testing, yielding outcomes much better than the 8% success rate seen in medicine making it to market.
To see similar results in your organization's processes, here are three best practices to keep in mind for successfully embracing AI:
• Existing Workflow Integration: Success with AI initiatives often breaks down when adopting the technology ends up breaking a company’s existing workflows. Needing to assemble new standalone teams, or reorganize existing teams, to implement an AI initiative doesn’t work. Focus on AI approaches that bring tools that integrate with existing systems and your organization’s existing expertise.
• Align On Value: As with any technology implementation, be sure the outcomes that prove an AI initiative’s value are clearly defined. Identify the benchmarks for value and make sure the initiative is directed at achieving those specific goals: time savings, cost reductions, prediction accuracy, etc.
• Generalization: Adopt an infrastructure that isn't purpose-built for a specific use case but that instead can be improved across many customer applications. This allows you to continue developing new assets rather than starting from scratch with every engagement.
Just as Yeager proved the impossible was possible, cross-functional teams of pharmaceutical scientists, software engineers, AI/ML and simulation experts are proving that drug development can be made faster, cheaper and with far greater efficiencies to better serve patients by many orders of magnitude. Your organization can do this, too, whether you're working in healthcare and drug development or innovating in another field.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
