A major review led by the University of Warwick reframes the question behind brain digital twins: the bottleneck isn’t computing power, but how much of a living, individual brain researchers can observe, validate and continuously update.
The idea of a “digital twin” of the human brain — a software model that mirrors a specific person’s brain closely enough to predict how it will respond to disease, injury or treatment — has become one of neuroscience’s more ambitious long-term goals. A review published in Nature Reviews Electrical Engineering and led by Dr Ruohan Zhang of the University of Warwick’s Warwick Manufacturing Group (WMG), reported on 2 September 2026, argues that the field has been asking the wrong central question. Rather than “how many brain cells can we simulate,” Zhang and colleagues argue the operative question should be “how much of an individual living brain can we actually observe, constrain and update in order to create a true digital twin.”
That reframing matters because it shifts the perceived bottleneck. Public discussion of large-scale brain simulation has tended to treat computing power and simulated neuron count as the limiting factors — a framing borrowed, understandably, from the trajectory of large language models, where scale has driven much of the recent progress. Zhang’s review argues that a digital twin brain is fundamentally different from a generic large-scale simulation: its defining feature is that it represents one specific individual and remains connected, on an ongoing basis, to that person’s biological brain through real measurement data. A model that is not continuously validated and updated against real observations of the actual brain it claims to represent is not a digital twin in the sense the field is aiming for, however many neurons it simulates.
A three-stage roadmap
The review sets out what its authors describe as a path from where the science currently stands towards where digital twin brains are headed. The first stages involve reconstructing brain structure as faithfully as current measurement technology allows — a task made harder by the sheer scale mismatch between what can currently be observed (limited windows into neural activity, structural imaging with finite resolution) and the brain’s actual complexity, which runs to tens of billions of neurons and vastly more synaptic connections. A further, more advanced stage envisions the digital twin and the biological brain responding to the same real-world inputs in parallel — watching, predicting, acting and adjusting together — and, further still, a stage where sustained interaction lets both digital and biological sides reshape each other over time, in a manner echoing how learning consolidates in a real brain. The review is explicit that current science sits at the earlier stages of this roadmap, with the more interactive, co-evolving stages remaining a longer-term prospect rather than a near-term deliverable.
Why the measurement framing matters
The practical implication of Zhang’s argument is that progress in digital twin brains will likely track advances in neural measurement technology — better, less invasive ways of recording activity from more of the brain over longer periods — at least as closely as it tracks advances in computing hardware or model architecture. This is a meaningfully different research priority than simply building bigger simulations, and it has implications for where funding and engineering effort are best directed: a digital twin built on sparse, infrequently updated measurements of a real brain, however computationally sophisticated its underlying model, will remain a poor twin of the specific individual it is meant to represent.
Why it matters
Digital twin brains, if achieved, carry substantial potential clinical value: testing how an individual patient’s brain might respond to a specific drug, surgical intervention or neurostimulation protocol in software before attempting it on the actual patient, or tracking neurodegenerative disease progression against a continuously updated personal baseline rather than population averages. The review’s core claim — that the limiting factor is measurement rather than raw computing scale — is a useful corrective for a field where public expectations have sometimes run ahead of what current neural recording technology can support. It also has a practical bearing on how research funding and engineering effort should be allocated: continued investment in measurement technology (higher-resolution, longer-duration, less invasive ways of recording from living brains) may do more to advance genuinely personalised digital twin brains than further increases in raw simulation compute alone, a nuance relevant to India’s own growing neurotechnology and computational neuroscience research base as it decides where to direct resources in this space.
– Srinayana Kavuri
Key facts
- Review: “Building digital twin brains at the limits of measurement,” Nature Reviews Electrical Engineering, led by Dr Ruohan Zhang, WMG, University of Warwick
- Core argument: digital twin brain accuracy is limited by what can be measured, validated and updated from a living brain — not primarily by computing power or simulated neuron count
- Sets out a three-stage roadmap: structural reconstruction, real-time parallel response, and long-term co-evolving interaction between digital twin and biological brain
- Reported 2 September 2026
