GGR Newsletter
June 2025
GGR Newsletter
June 2025
Mary D. Cundiff, Ph.D.
June 2025
Wouldn’t it be great if we could personalize medicine? What if we could tailor treatment to your specific biology instead of relying on broad, one-size-fits-most approaches? It might sound a bit futuristic or too good to be true, but we might be closer to this reality than we realize.
Why Personalized Medicine Matters
Personalized medicine has been a topic of discussion for years. Most treatments today are based on what works best for most people. Appendicitis? Remove it. Common cold? Take some Nyquil. Headache? Try ibuprofen. Cancer? Use radiation and chemotherapy.
This approach works well for many conditions, but what about the edge cases? What about the people who don’t respond to standard treatments? Historically, it hasn’t been worth the time or cost to target the outlier populations. The reality is that every individual’s biology is different. However, thanks to the rise of computational biology and the explosion of biological data, we now have the tools to consider the unique biology of each person.
Treatment ≠ Cure
It’s important to remember that there is a big difference between a treatment and a cure. Much of medicine today focuses on managing symptoms, not reversing or halting disease progression. Cancer is a prime example. Cancer is incredibly complex, and current treatments like radiation and chemotherapy are blunt approaches. They don’t cure, they attack; often harming healthy cells in the process. It’s a matter of hoping the treatment beats the disease before the side effects beat you. Thankfully, we have been able to make great strides in creating technology to cure several diseases, treat metabolic disorders, and manage HIV. Despite significant research, we still have a lot to learn if we want to live in a world with more cures than treatments.
The Role of the Cell
Personalized medicine isn’t just for cancer though… it applies to all diseases. But to truly make it work, we have to start with the fundamentals of biology: the Cell. In a recent perspectives piece from Cell Press, researchers discuss how advancements in AI and omics make it possible to build models of cells directly from data. Thanks to these advancements, we have new opportunities to tackle the idea of an AI virtual cell (AIVC) that aims to not only represent but also simulate molecular, cellular, and tissue behavior across various states (Figure 1).
Current Capabilities
We already have some strong foundations:
Data Explosion: We now have massive datasets from transcriptomics, proteomics, and imaging, not only at a single-cell level but with growing details of spatial profiling. Even more so, major collaborations, like the Human Cell Atlas, CellScapes, and folding@home, have provided exceptionally large platforms for data availability.
AI Tools: With high-dimensional data, we have been able to use large neural networks like AlphaFold to learn biological patterns. Machine learning models have been wildly successful as they are predictive, generative, and queryable.
Thanks to existing biological models, simulation tools, and some major collaborations we seem well equipped to develop an AI virtual cell (Figure 2). So what exactly is standing in our way?
Challenges and Opportunities
Naturally, this complex endeavor of developing a virtual cell has been met with a multitude of challenges:
Multi-scale modeling
Cells operate on many levels. You have atoms, molecules, and tissues; where changes at one level can trigger large effects at another level. These relationships are not always straightforward. To simulate this, we need unified or universal representations that bridge information between sources like imaging and sequencing.
Diverse processes with many components
Cells have an insane amount of functions that result from complex molecular interactions. To name a few main characters, you have:
Gene regulation, or turning on and off genes based on the context.
Metabolic pathways, or processing energy.
Signal transduction, or converting signals (like hormones) into cellular responses.
Modeling these systems requires capturing interactions among thousands of molecules and their ever-changing configurations.
Nonlinear dynamics
Biology isn’t always predictable. A tiny change can have a big and sometimes unexpected effect, because the systems do not behave in a simple way. We need dynamic models that can simulate the complexity of evolving, perturbed states.
Interpretability
There is a delicate balance between deep learning “black boxes” and mechanistic understanding. To get scientific value, we need causal and actionable models, helping design targeted experiments. Rather than simply finding correlations within the data, we need the models to be understandable, even to non-experts.
Virtual Instruments and Evaluation
Once we have built universal representations of our biological components, we need to be able to simulate experiments or treatments. Virtual instruments are tools that can manipulate or decode cell behavior. They enable active learning and the self-improvement of our models.
Collaborative and Ethical Development
This project is definitely not a solo effort. It will require multidisciplinary teams across biology, AI, ethics, and policy. Biological tools are a hot topic for ethical discussion. CRISPR is a great example of a biological tool that was designed for research purposes but has shown great potential and warrants regulation.
A Future Within Reach
While many challenges lie ahead, the building blocks are already falling into place. The progress we have made as a scientific community is a massive accomplishment. Personalized medicine isn’t science fiction anymore. With continued research, ethical development, and global collaboration, a future where medicine is tailored to you may be closer than we think.