A single breakthrough idea in computational modeling could one day change how we understand – and possibly repair – human vision. And this is the part most people miss: it’s not a new drug or a surgery, but a virtual retina built inside a computer.
Researchers at the University of Surrey have developed a pioneering computer model that simulates how the retina forms and regenerates, potentially paving the way for new strategies to treat vision loss.
What this new model actually does
In simple terms, the team has created a digital version of retinal development that shows how a complex, layered eye structure can grow from just one kind of stem cell.
Instead of looking at cells under a microscope alone, they use mathematical and computational rules to follow how identical starting cells gradually turn into the many specialized cells needed for sight.
This gives scientists a clearer, step-by-step picture of how vision emerges during development and how that process might go wrong in disease or after injury.
Simulating retinogenesis step by step
The model focuses on retinogenesis – the process where early, identical progenitor cells diversify into the six major types of neurons that make up the retina.
By using an approach called agent-based modeling, each cell in the simulation behaves like an individual “agent” that can grow, divide, move, and change identity based on built‑in rules.
This lets the researchers replay key stages of retinal formation on a computer, watching how a uniform cell population turns into a highly organized neural tissue.
Simple rules, precise retinal layers
One striking outcome of the work is that relatively simple genetic rules, plus a bit of randomness, are enough to recreate the retina’s finely layered architecture.
Those layers are not just decorative; their precise arrangement is essential for how light signals are processed and ultimately turned into vision.
This raises a bold question: if such order can emerge from simple rules and chance, how much of eye development is “hard‑wired” and how much is flexible – and could that flexibility be leveraged in regeneration?
Where the research was shared
The team presented their findings at the IWWBIO 2025 conference, a venue that brings together experts in bioinformatics and computational biology.
Their work is also published in the Lecture Notes in Computer Science (LNCS) series, giving other researchers access to the methods and results so they can build on or challenge them.
That visibility matters, because controversial or unexpected modeling results only become useful when the wider community can test and debate them.
A closer look at the biology insight
Lead researcher Cayla Harris from the University of Surrey’s Nature Inspired Computing and Engineering Group emphasizes a powerful idea: in biology, extremely complex structures often arise from surprisingly simple underlying rules.
In this project, the simulations show how genetically identical cells, influenced by small internal biases and random events, can self‑organize into the retina’s neatly ordered layers.
This challenges the intuitive belief that every detail must be precisely pre-programmed and instead suggests that robust patterns can emerge from a mix of guidance and chance.
How the virtual retina is built
To construct this digital retina, the team used the BioDynaMo software platform to model thousands of virtual cells.
Each cell in the simulation can grow, divide, and make “fate decisions” about what kind of neuron to become, based on an internal gene‑regulation logic that mimics biological networks.
By tweaking how these virtual genes interact, the researchers can test different hypotheses about what actually drives real cells to choose one identity over another.
Competing gene network designs
The study compared several possible designs for the gene networks that control cell fate decisions during retinal development.
Two specific configurations, called the Reentry model and the Multidirectional model, matched real biological data most closely when used in the simulations.
This suggests that, instead of following a strict one‑way sequence, retinal cells may choose their fate through overlapping, flexible pathways where multiple options remain available for longer than many textbooks imply.
A potentially controversial interpretation
If retinal cells really do rely on overlapping and adaptable genetic pathways, that could challenge older, more rigid views of how neural tissues develop.
Instead of a fixed “decision tree” where each step permanently closes off alternatives, development might behave more like a dynamic network where cells keep several options open before settling.
Some scientists might argue this adds unnecessary complexity to models of development, while others might see it as a more realistic reflection of biology’s built‑in flexibility.
What do you think: should models prioritize simplicity, or is it worth embracing messy, flexible systems if they better match reality?
Why this matters for disease and repair
Because the model can simulate normal retinal development, it also offers a framework to explore what happens when things go wrong.
Researchers can, in principle, introduce virtual “mutations” or disruptions to see how diseases of the retina might arise from altered cell decisions.
The same approach could inform regenerative medicine, where scientists try to guide stem cells to rebuild damaged retinal tissue in conditions like macular degeneration or inherited retinal disorders.
The power of computational experiments
Computational modeling offers a way to run experiments that are difficult, slow, or impossible to perform directly in living tissue.
By tracking every simulated decision a cell makes and every interaction it has with its neighbors, scientists can test many “what if” scenarios quickly and safely.
That makes models like this valuable not as final answers, but as tools to generate new, testable predictions for laboratory and clinical research.
Who supported the work
The project received support from the UK’s Engineering and Physical Sciences Research Council (EPSRC), a major funder of research at the intersection of mathematics, computing, and physical sciences.
This kind of backing reflects a broader trend: funding agencies increasingly recognize that solving biological and medical problems often requires tight collaboration between biologists, computer scientists, and engineers.
Linking genes, code, and development
Cayla Harris and colleagues see their work as one step toward uniting genetics, computational methods, and developmental biology around a single, shared target: understanding how one of the body’s most intricate neural structures takes shape.
The retina is often described as an accessible part of the brain, and modeling its development could shed light on broader principles that apply to other neural tissues as well.
If this approach proves robust, similar agent‑based models might eventually be used to explore development and repair in the cortex, spinal cord, or other parts of the nervous system.
Reference and further reading
The main study is reported as: Harris, C., et al. (2025). Agent-Based Modelling of Retinal Development, published in the Lecture Notes in Computer Science series, doi: 10.1007/978-3-032-08452-1_6.
Interested readers can consult the publisher’s website for the full chapter and technical details.
Important notes and disclaimers
This kind of research deals with basic science and computational modeling; it does not yet translate into approved medical treatments.
Nothing here should be taken as medical advice or as a recommendation to change or start any treatment.
Anyone seeking information on eye health, retinal disease, or vision loss should always consult a qualified medical professional.
Also, when interacting with online services to ask research or health‑related questions, avoid sharing sensitive or confidential personal data.
Always review the platform’s terms, privacy policies, and data‑handling practices before submitting questions.
And here’s where it could get really controversial… If a computer model can one day predict how to rebuild a damaged retina, should that model carry as much weight in decision‑making as traditional lab experiments?
Would you personally trust a treatment strategy that was heavily guided by simulations like this, or do you feel biology should stay in the hands of “wet lab” data only?
Share where you stand: are you excited, skeptical, or somewhere in between?
