The Changing Role Of Computational Mechanics In Drug Discovery

By Amanda Randles, Ph.D., Director, Duke University’s Center for Computational and Digital Health Innovation; Head of the Randles Lab

Metastasis remains responsible for more than 90% of cancer deaths despite remarkable advances in molecularly targeted therapies. Over the past several decades, cancer drug discovery has been transformed by genomics, proteomics, and high-throughput screening, leading to an increasingly detailed understanding of the molecular pathways that drive disease. These advances have identified new therapeutic targets and produced more effective precision medicines. They have also reinforced an important realization: understanding the molecular biology of a tumor is only part of the challenge.

The metastatic cascade is as much a mechanical process as it is a biological one. Before a circulating tumor cell can establish a secondary lesion, it must survive transport through the bloodstream, deform through vessels that may be smaller than the cell itself, interact with red blood cells and platelets, adhere to the vessel wall under flow, and ultimately cross the endothelium into surrounding tissue. Every step depends on molecular signaling, but every step is also governed by mechanics. Cell deformability, blood flow, vascular geometry, and the surrounding cellular environment all influence whether metastatic spread succeeds. Experimental and computational studies over the past decade have made it increasingly clear that these physical interactions are active regulators of metastatic progression, not simply consequences of disease.^1–3

Experimental systems, such as microfluidics, organ-on-a-chip technologies, and engineered vascular models, have transformed our ability to study these processes under physiologically relevant conditions. Yet they also illustrate one of the central challenges of metastasis research: Biology and mechanics are tightly coupled. Altering membrane stiffness may also change cytoskeletal organization. Changing receptor expression can influence both adhesion and cell mechanics. Even carefully controlled experiments inevitably produce different cellular configurations, collision histories, and flow conditions, making it difficult to isolate the contribution of any individual mechanism.

For decades, computational models have helped interpret experimental observations and improve our understanding of metastatic transport. I believe their role in drug discovery is beginning to change. As these models become increasingly physiologically realistic and computationally scalable, they are evolving from tools that explain biology into platforms that can help design experiments and prioritize competing hypotheses before they are tested experimentally. Computational models will never replace experimental biology, nor should they. Their emerging role is different. They can help us decide which experiments are most likely to reveal the biology that matters.

Computational Experiments

One of the greatest strengths of experimental biology is that it captures the complexity of living systems. That same complexity, however, makes it remarkably difficult to isolate the contribution of any individual mechanism. Metastasis illustrates this challenge particularly well. A circulating tumor cell may successfully adhere because it expresses more adhesion receptors, its membrane is more deformable, its nucleus is stiffer, its collisions with surrounding red blood cells alter its trajectory, or local blood flow changes the biology of the endothelium it encounters. In reality, all of these processes occur simultaneously and influence one another. As a result, experiments often demonstrate that several mechanisms contribute to metastatic progression without clearly revealing which has the greatest influence.^1–3

Computational models offer a different way to approach this problem. Rather than asking whether a simulation reproduces an experimental observation, we can begin with a specific biological hypothesis and ask what happens when only that mechanism changes. The vascular geometry remains the same. Blood flow remains the same. Every surrounding red blood cell follows the same trajectory and undergoes the same deformation. We then modify a single variable. The nucleus becomes stiffer. Membrane elasticity changes. Receptor expression increases. Because every other aspect of the system is held constant, the contribution of that individual mechanism can be measured directly. This level of experimental control is extraordinarily difficult to achieve in vitro or in vivo.

Advances in computational science have made these studies increasingly realistic. Modern simulations now combine physiologically realistic blood flow with deformable models of red blood cells, circulating tumor cells, platelets, and the vascular wall, allowing individual cellular interactions to be studied within tissue-scale vascular networks.⁴^,⁵ At the same time, advances in algorithms and high-performance computing have made it possible to perform these simulations at scales that were not feasible only a few years ago.⁶

To me, that is what is changing. Computational models are no longer limited to explaining observations after experiments have been completed. They are beginning to help determine which experiments should be performed first. By comparing competing biological and mechanical hypotheses under identical physiological conditions, computational models can help identify the mechanisms most likely to drive disease and focus experimental effort where it is most likely to have the greatest impact.

Expanding The Therapeutic Search Space

Perhaps the most important consequence of computational experiments is that they expand where we look for therapeutic opportunities. Drug discovery has traditionally focused on identifying molecular targets, and for good reason. Molecular signaling underlies nearly every aspect of cancer progression. Metastasis, however, is both a biological and a physical process. A circulating tumor cell must survive transport through the bloodstream, deform through confined vessels, interact with surrounding blood cells, approach the vessel wall, form adhesive bonds, and ultimately cross the endothelium. Each of these steps depends on molecular biology, but each is also shaped by mechanics. Once metastasis is viewed through both lenses, the search for therapeutic intervention naturally broadens. Instead of asking only which molecule should be targeted, we can begin asking which biological or mechanical process most limits metastatic success.

One opportunity lies in the physical properties of the tumor cell itself. Membrane elasticity, cortical tension, and nuclear stiffness all influence how circulating tumor cells deform, migrate, and interact with their surrounding environment. Experimental studies have shown that metastatic cells often exhibit mechanical behavior distinct from less aggressive cell types, but determining which of these properties actually drives metastatic progression remains challenging because they are so tightly coupled.^4,7 Using experimentally derived mechanical measurements within physiologically realistic simulations, we have shown that individual cancer cell lines can be represented directly in computational models of blood flow.⁴ Rather than treating deformability as another characteristic to measure, these models allow us to ask whether changing a specific mechanical property meaningfully alters transport, adhesion, or extravasation. If it does, that property becomes more than an interesting biological observation. It becomes a potential therapeutic target.

Mechanics alone, however, is unlikely to explain metastatic behavior. Adhesion ultimately depends on molecular interactions between circulating tumor cells and the vascular endothelium. Receptor expression, ligand availability, and bond kinetics all contribute to whether a cell remains attached long enough to extravasate. Computational models provide the ability to evaluate these molecular mechanisms within a realistic mechanical environment rather than in isolation. If receptor expression proves to dominate adhesion under physiological flow, then molecular therapies remain the logical direction. If relatively modest changes in cell mechanics produce larger changes in adhesion than substantial changes in receptor density, that finding shifts experimental attention toward a different class of interventions. Rather than replacing traditional approaches to target discovery, computational models provide another layer of evidence for prioritizing competing biological mechanisms before committing to extensive experimental studies.

Perhaps the most intriguing opportunity lies within the vessel wall itself. Endothelial cells continuously sense and respond to the mechanical forces generated by flowing blood. Wall shear stress influences inflammatory signaling, endothelial activation, and the presentation of adhesive molecules, suggesting that blood flow does more than transport circulating tumor cells. It also helps determine the biological environment they encounter.⁸ In our recent work, we developed a multiscale framework that couples tissue-scale blood flow with receptor ligand adhesion whose kinetics depend on local wall shear stress.⁵ Rather than assuming adhesive behavior is constant throughout the vasculature, the model allows endothelial receptor activity to vary according to the local hemodynamic environment. The resulting simulations predict striking spatial differences in circulating tumor cell arrest within the same vascular network, suggesting that regional flow conditions may actively regulate where metastases are most likely to develop. This expands the search space even further. Instead of focusing exclusively on the tumor cell, we can begin asking whether modifying endothelial mechanobiology or the vascular response to flow might reduce metastatic seeding before tumor cells ever leave the bloodstream.

Integrating Computational Mechanics Into The Drug Discovery Pipeline

Drug discovery has become increasingly multidisciplinary. Advances in genomics, imaging, single-cell sequencing, organ-on-a-chip technology, and artificial intelligence have each expanded the types of biological questions we can ask. Rather than replacing one another, these approaches provide complementary views of disease. Genomics identifies molecular alterations. Imaging reveals anatomy and tissue organization. Organ-on-a-chip systems recreate physiological environments under controlled conditions. Computational mechanics contributes something different. It provides a physiological framework for understanding how physical forces influence cellular behavior within those biological systems.⁹

The greatest opportunity, in my view, is to integrate computational mechanics into the earliest stages of therapeutic discovery. Consider a patient with metastatic disease. Medical imaging defines the vascular anatomy. Experimental measurements characterize the mechanical properties of circulating tumor cells. Molecular profiling identifies candidate receptors and signaling pathways. Rather than pursuing every plausible mechanism experimentally, computational models can integrate these data into a common physiological framework and evaluate competing biological hypotheses. Which mechanism most strongly influences transport? Which dominates adhesion? Which is most likely to determine whether metastatic seeding succeeds? Those questions can begin to be addressed before the first follow-up experiment is designed.

Viewed this way, computational mechanics becomes less of a validation tool and more of a decision-making tool. Experimental systems provide the biological measurements needed to build realistic computational models. Computational models help identify the mechanisms that warrant deeper experimental investigation. Those experiments then refine the models, creating an iterative cycle in which computation and experimentation continually strengthen one another. Rather than existing as separate approaches, they become components of a single discovery process.

I do not envision a future in which drug discovery becomes fully computational, nor do I think it should. Biology will always surprise us, and every computational prediction must ultimately be tested experimentally. What I believe is beginning to change is where computational mechanics contributes. For much of the past several decades, computational models have helped explain biology after experiments were completed. Increasingly, they are helping determine which experiments should be performed first. If they can reliably narrow a broad hypothesis space into a smaller set of experiments with the greatest potential to reveal meaningful biology, their greatest contribution may not be predicting disease. It may be helping us decide where discovery should begin.

Looking Ahead: Computational Mechanics As A Discovery Platform

The implications of this approach extend well beyond metastasis. The same computational frameworks developed to study circulating tumor cells can be applied to nanoparticle delivery, engineered immune cells, cell therapies, and other biologics whose success depends on transport through the vasculature. As therapies become increasingly personalized, understanding how they move through the body may become just as important as understanding how they interact with their molecular targets.

This evolution has been driven as much by advances in computational science as by advances in computing hardware. Modern algorithms now make it possible to bridge scales ranging from receptor ligand interactions to blood flow throughout complex vascular networks. In our own work, approaches such as adaptive physics refinement have made these multiscale simulations computationally practical, enabling entirely new classes of computational experiments.⁶

Drug discovery has always depended on asking the right questions. Computational mechanics is beginning to provide a rigorous framework for comparing competing biological and mechanical hypotheses before committing years of experimental effort. Its greatest contribution may not be predicting biology. It may be helping researchers decide where discovery should begin.

References

1. Wirtz D, Konstantopoulos K, Searson PC. The physics of cancer: the role of physical interactions and mechanical forces in metastasis. Nat Rev Cancer. 2011. 
2. Au SH et al. Clusters of circulating tumor cells traverse capillary sized vessels. PNAS. 2016. 
3. Follain G et al. Hemodynamic forces tune the arrest, adhesion, and extravasation of circulating tumor cells. Dev Cell. 2018. 
4. Balogh P, Martin CM, Randles A, et al. Data driven modeling of circulating tumor cell mechanics in physiologically realistic blood flow. Sci Rep. 2021. 
5. Martin A, et al. Multiscale Modeling of Shear Dependent Tumor Cell Adhesion. ICCS. 2026. 
6. Roychowdhury S, et al. Enhancing adaptive physics refinement simulations through the addition of realistic red blood cell counts. SC23. 2023.
7. Xu W, Mezencev R, Kim B, et al.
   Cell stiffness is a biomarker of the metastatic potential of ovarian cancer cells.
   PLoS One. 2012.
8. Davies PF. Hemodynamic shear stress and the endothelium in cardiovascular pathophysiology. Nature Clinical Practice Cardiovascular Medicine. 2009.
9. Huh D, Hamilton GA, Ingber DE. 3D cell culture to organs on chips. Trends in Cell Biology. 2011.

About The Author

Amanda Randles, Ph.D., is a professor of biomedical engineering at Duke University, where she directs the Duke Center for Computational and Digital Health Innovation. Her research develops patient-specific digital twins that combine high-performance computing, artificial intelligence, and multiscale biophysical simulation to enable earlier diagnosis, personalized treatment, and improved understanding of diseases ranging from cardiovascular disease to cancer. Her work has been recognized with the ACM Prize in Computing, the NIH Director's Pioneer Award, the NSF CAREER Award, the ACM Grace Hopper Award, the Jack Dongarra Early Career Award, and the inaugural Sony–Nature Women in Technology Award. Randles received her Ph.D. in applied physics, M.S. in computer science from Harvard University, and B.A. in computer science and physics from Duke University.