One of the most challenging aspects of treating many diseases, including cancer and inflammatory conditions, is that they are highly heterogeneous. In patients with the same diagnosis, both clinical presentation and the underlying mechanisms of the disease can diverge significantly. Even identical clinical presentation or medical imaging do not necessarily reflect the same underlying pathogenesis. The result is that drug treatments often work only for some patient groups, and drug nonresponse can even worsen the course of the disease or its symptoms in others.
As knowledge of the heterogeneous nature of diseases has increased exponentially, drug development has indeed become more sophisticated. With more data available about the molecular nature of disease, possible targets and the molecules that can treat those conditions, the future is more optimistic than ever. But drug development has also become more expensive, with greater risk when new drugs fail to prove safe and effective in clinical trials.
The average cost of developing a drug increased fivefold between 2000 and 2018, including failures, while the ratio of R&D spending to sales rose from 11.9% to 17.7% over the same period, according to one of the most comprehensive studies on the topic.
Moving efficiently and successfully from the molecular development stage to the clinical stage remains a formidable challenge.
However, the emergence of computational disease models, molecular maps of disease which consider a nearly infinite amount of data and parameters in pharmaceutical R&D, are proving an effective tool to manage the challenge. As a gastroenterologist, I see cell-centered computational disease models as finally offering a way to both gain insight into disease pathogenesis and medical treatments and allow medical needs to integrate into considerations driving business interests.
Only between 29% and 34% of Phase II drug trials succeed. This means that most fail, and do so after many years of research and investment. In fact, only about 10% of compounds developed by pharma labs ever make it to market. Among the largest 18 pharmaceutical companies, the rate of likely approval for any compound is just 14%. Much of this is due to lack of sufficient accurate translation of the target’s biological effects into the disease space, with it often taking years for the errors (or success) to show themselves.
Computational disease models, enriched by LLMs’ insights, are starting to change this. These engines can more accurately predict which targets are the most valuable or effective and in which disease they have a higher chance of success. Considering the varied nature of disease, they can help select different targets for the same diseases, and identify a better pool of patients to test efficacy, leading to more efficient trials. They can accomplish in minutes the computational work that would take scientists years.
Another challenge is that single molecules offer only partial coverage of the patient population in each disease; combination therapy is considered a valuable solution for this issue. Computational disease models can help determine how different combinations of drugs can help patients in different disease subpopulations and with different clinical manifestations. Currently, combinations take into account known pharmacodynamic drug properties and parameters such as time to induction of drug effects. However, computational disease models can predict with more mechanistic precision what combinations would work for each type of patient.
These models will not only drive science. They will allow pharma companies to invest more efficiently, ultimately leading to lower costs and more medication availability for everyone. Currently, business interests are the leading cause of halted clinical trials, accounting for 36% of terminated trials, versus 24% for lack of efficacy. Pharma companies often end trials because of fears of competition, indications that the compound would be better suited for a different condition or due to mergers and acquisitions sparking a realignment of priorities. Because of how fast they work and the number of parameters they can integrate, computational disease models have the potential to transform indication prioritization. Pharma companies will be able to more quickly determine on which disease to focus, how to proceed to the trial phase and which patients to include for the best results.
The in-depth, constantly-changing birds-eye view of disease that computational models provide is an essential next step in linking our ever-expanding clinical knowledge and data with drug development. It will increasingly focus and derisk trials, giving the pharma industry a path toward helping more patients at lower cost.
Photo: pagadesign, Getty Images
Yehuda Chowers, MD, is Chief Medical Officer at CytoReason. With over 30 years of experience in medical practice and research, Prof. Chowers is one of Israel’s leading gastroenterologists. He previously served as Director of both the Research Division and the Gastroenterology Institute at Rambam Hospital, and as Professor in the Technion Faculty of Medicine.
This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.
