Two powerful forces behind three product launches in 2022: Mechanistic modeling & brainpower

Our science & innovation

Scientist notes

Research and development

Two powerful forces behind three product launches: Mechanistic modeling and brain power

The future of pharmaceutical research is already here at Bristol Myers Squibb.

Three unique drugs launched by the company in 2022 are proof that new approaches can accelerate results.

Those medicines target three different diseases — melanoma; hypertrophic cardiomyopathy; and plaque psoriasis — and yet, they share one important trait.

All were developed with the help of mechanistic modeling, using a suite of mathematical techniques to analyze and make predictions about complex biological systems.

A brainpower-driven approach

The world has been captivated by a different type of predictive technology: automated methods such as artificial intelligence (AI) and machine learning (ML), which make calculations by detecting statistical correlations in large quantities of data. Those techniques have made remarkable leaps, including their ability to identify drug candidates, but they can only perform the tasks for which they’ve been programmed.

Like AI and ML, mechanistic modeling can greatly accelerate the process of drug development, providing insights that once would have required clinical trials.

But unlike those technologies, mechanistic modeling focuses on the causal relationships between intricate sets of variables. Researchers create a mathematical model to represent the underlying mechanics of the system being studied, incorporating the laws of natural science. This computer-based model is then used to generate and test hypotheses about how the system will behave under varying conditions.  

Mechanistic modeling is indispensable for tasks in which the “why” is crucial — for example, using knowledge about a disease’s molecular pathways to predict the range of drug dosages likely to achieve optimal results across different patient populations.

Teams leverage sharp, sophisticated modeling skills to save time

At BMS, multidisciplinary “matrix teams” use cutting-edge modeling techniques such as quantitative systems pharmacology and physiologically based pharmacokinetics to inform decision-making, optimize and streamline workflows, and shorten the development cycle.

“The process is driven by expert mechanistic modelers – extremely gifted people with a good understanding of biology, mathematics and statistics,” said Akintunde Bello, senior vice president, Clinical Pharmacology, Pharmacometrics, Disposition and Bioanalysis. “It starts with an exhaustive search of the literature around the disease pathways we’re interested in. What are the cellular components involved? How do they interact?”

Multiple steps, tests involved

The modelers work with our biologists, translational science experts, clinical trial investigators and clinical pharmacology leads to gather additional data and deepen their understanding of the problem under study.

They use ordinary differential equations to characterize each mechanism mathematically, then build models of complex systems by linking those equations together – sometimes hundreds at a time.

Next comes the validation phase, in which the model is tested in different simulated scenarios to see if it successfully predicts the outcome. Refinements may follow, increasing the model’s accuracy.

“Mechanistic modeling can have different flavors, from extensive disease models down to how your drug interacts with a target receptor,” Bello noted. “Quantitative systems pharmacology, for example, can be useful in determining how much drug you need on board to get maximum benefit and minimal toxicity, and thus help in selecting an optimal dosing regimen.”

How the challenges can reap great rewards

These methods also pose some challenges. “You need to have a strong idea of the biology that you’re trying to model,” Bello explained. “We have significant expertise and emphasis on understanding and addressing root-cause disease biology, however, there’s often some missing information, which can add uncertainty to the models.”

Mechanistic modeling also requires patience. “It can take up to a year to develop some of the more complex disease models – though that’s still much shorter than running most clinical trials,” he said. “And talent acquisition isn’t easy. Good modelers are in very strong demand right now. There’s fierce competition for their skills.”

Ultimately, though, the promise of such techniques is enormous. “What excites me is the potential impact these models can have on increasing the quality of the assets we decide to progress, while reducing development time,” he said. “Once they’re built, they can be modified and utilized again and again, for a wide range of purposes. They’re a powerful gift that keeps on giving.”