Mark Eller, PhD, is senior vice-president (SVP) of Research and Development at Aria Pharmaceuticals. Eller has held several vice-president/SVP roles leading teams from discovery through approval at large and mid-sized pharma companies, including Jazz, Quintiles (now IQVIA), and HMR (now Sanofi). Eller is co-inventor on 30 patents and has led health authority presentations and negotiations, including six FDA advisory committee meetings. He has co-authored numerous scientific publications and served as an adjunct professor of Pharmaceutics at the University of Cincinnati. Eller received his Pharmacy degree and PhD in Pharmaceutics from the University of Iowa.

Mark Eller

Scientists and researchers have made remarkable breakthroughs in basic science to better understand the complex pathology of many diseases today. Some of that progress is through utilizing technology to uncover new insights, and some of the progress is through the incalculable capacity of the human mind to analyze problems and find solutions. With all that knowledge and all the effort put in, however, human biology and disease pathology continue to be surprising. Every day researchers continue to make profound discoveries in basic science that create new possibilities in fighting disease.

A conundrum exists, however: the more knowledge that is accumulated about a disease, the harder it becomes to incorporate all that data to understand the disease complexity, and, ultimately, the more difficult it is to find treatments. Complex diseases require complex solutions to untangle the disease biology because human biology is interconnected — millions of signals and cellular interactions are taking place each second. That said, a better understanding of that complexity also creates an opportunity to attack the disease in different ways and explore new insights that could treat or cure many of today’s complicated and hard-to-treat diseases.

To untangle that disease complexity, scientists are using technology to their advantage. Modern computational approaches can allow them to analyze disease data collected over decades in ways that were never previously imagined. A significant drive has been placed on artificial intelligence (AI) as a supplement to how scientists take that data and screen libraries of drug candidates against it to find new potential therapies.

Many companies are doing just this, but what they are effectively achieving is an acceleration of the traditional research and discovery process. At its simplest form, the traditional drug discovery process involves taking a single disease hypothesis, such as a widely known or novel disease target or the phenotype of a specific disease, and trying to develop or discover new molecules that can modulate it. Computers utilizing AI can do this more efficiently, enabling researchers to accelerate screening efforts for new therapies. There is little debate about the utility of AI here.

However, a fundamental problem remains. Researchers are targeting diseases through one or two known characteristics of that disease, whether it’s a specific receptor or a more complex disease pathway. The researchers can target several receptors, for example, but currently the industry is doing that in a silo. Typically, the researcher targets a specific receptor, gathers insights and data, then targets another receptor. This activity continues on down the line until the researcher understands how a molecule or drug interacts with that disease.

Imagine instead, however, targeting the whole disease and understanding how any particular molecule or drug simultaneously modulates all those receptors. Or better yet, building a complete model of a disease and analyzing a potential molecule or treatment against the complex inner workings of that disease, from receptors to ligands to other disease signals. The only true way to do that definitively is through human studies, but the efficacy and safety of therapies must first be explored before even approaching clinical research. That early exploration is the phase where disruptive technology can have a significant impact and bring more potential treatments to the clinic.

It is at that early exploratory phase where the value of AI prevails. The next wave of innovation in AI must examine how all the puzzle pieces of human biology are connected, especially as scientists uncover more and more about disease pathology. This level of drug discovery has the potential to uncover new therapies and new mechanisms of action that target disease more safely and effectively, and this breakthrough is possible today.

In recent years, a team working with AI has been creating 

 models of a disease with dozens of heterogenous data types that can be used to screen databases of drug and drug-like molecules against that disease. Effectively, the team has been looking at how any given molecule will interact with human disease biology from dozens of different angles at the same time. This work is the closest to mimicking human trials in early preclinical research that has ever been observed, and it is work that could not be accomplished without the use of technology — in particular, AI.

Two critical lessons stand out in the team’s work. The first is that, by building these models, the researchers are not only rediscovering molecules that are known to work, have been tested, and passed human trials, but discovering completely new potential drugs that rank high in efficacy and safety that would otherwise not be found using any single-disease signal. The team’s work is a testament to the vast complexity of human biology and disease pathology. It also showcases the fact that potential treatments are being missed when only one dimension of a disease is examined against potential molecules.

The second lesson that stands out is that, if molecules or drugs can be tested through this approach, then scientists can gain much insight into how those molecules or drugs may work in human trials long before reaching the human clinical trial phase. This accomplishes several important goals, chief among those is de-risking experimental R&D. R&D is an extremely risky business, and researchers have come to accept that, for every successful treatment, billions of dollars will have been spent on thousands of other molecules that fail. Using technology to better model a more complete disease can help to pull those numbers in the right direction. There is no anticipated time for when technology will completely remove risk, but any progress in that direction is good progress.

The above lessons are just two examples of benefits that the bio/pharm industry can seek in drug discovery and development by harnessing technology to understand a more complete picture of any given disease. One thing is for certain, however: diseases will continue to be elusive and unpredictable. For every disease that researchers have found a treatment for, there exists millions of disparate data points. The more that scientists can leverage technology and AI to better analyze those pieces of the puzzle and generate a better picture of the whole disease, the better they will be able to identify and test potential therapies against a more holistic disease hypothesis.

Every barrier and problem that the bio/pharma industry faces has been met with new innovations to address them, and the growing complexity of disease data is no different. The potential of AI and other technologies to give a better understanding of diseases and how any given molecule or drug may affect that disease will be groundbreaking in drug discovery. In a world where knowledge of diseases becomes more complex, the technology could evolve to unravel that complexity. Thus, instead of finding fewer breakthroughs, the industry may see a continued steady rate of new discoveries despite having to deal with exponentially harder problems.

Mark Eller, PhD, is senior vice-president of Research and Development at Aria Pharmaceuticals.

Articles

Advancing the use of AI to understand the whole of a disease can reveal drug development insights that lead to drug discovery breakthroughs.