Alzheimer’s disease may not be a single, uniform condition, but a complex disorder shaped by different combinations of cognitive symptoms, brain changes, biology, genetics, and risk factors. Researchers propose using advanced computational modelling to bring these different layers together and build a more complete profile of each patient. The approach could eventually support more personalized diagnosis, better patient stratification, and treatments tailored to the processes driving disease in each individual.

For more than a century, Alzheimer’s disease has largely been discussed as a single condition, despite its complex nature. A new paper in Brain led by Prof. Shahar Arzy (Faculty of Medicine) Dr. Bark Raveh (Computer science) and Prof Gad Marshall (Harvard Medical School) suggests that novel computational approach may enable doctors to decipher the disease’s complex nature and consequently fit preventive interventions to the individual subject long before symptoms appear. Rather than searching for a single explanation that applies equally to every patient, the computational era may allow researchers to embrace the disease’s complexity, identifying the different pathways that can lead to cognitive decline and ultimately moving toward diagnosis and treatment tailored to the individual.

The researchers point out that people diagnosed with Alzheimer’s do not necessarily experience the same symptoms, show the same patterns of changes in the brain, or share the same biological and genetic basis. In some people, memory loss may dominate, while others may experience problems with orientation, language, behaviour, visual processing, or executive functions. At the same time, the biological processes associated with Alzheimer’s extend beyond the two proteins most commonly linked to the disease, amyloid and tau. There are important contributions of inflammatory, vascular, metabolic, infectious, toxic and other brain pathologies.

“This means that two people who receive the same diagnosis of Alzheimer’s disease may actually be experiencing very different combinations of biological and cognitive changes,” said Prof. Shahar Arzy. “Instead of looking for a single explanation that fits every patient, we need to understand the individual profile of the disease. The computational tools now available give us an opportunity to bring these different pieces together and ultimately move toward more personalized diagnosis and treatment.”

From one diagnosis to an individual profile
Rather than asking only whether a person “has Alzheimer’s,” the researchers propose building a much more detailed picture of each patient.
Such a profile could combine information about cognitive and behavioural symptoms with brain imaging, biomarkers, genetics, molecular data, risk factors, and the way the disease changes over time.

The goal is similar to the shift toward precision medicine already taking place in other areas of healthcare: understanding which processes are driving the disease in a particular person.

To make that possible, the researchers outline a computational approach known as metamodeling.

Instead of trying to building from scratch a new single enormous computer model and squeeze every piece of information about Alzheimer’s into it, the approach makes use of existing (and newly built) separate, more specialized models, each examines a different part(s) of the disease. One might focus on brain imaging, another on genetics, another on specific molecular changes, and another on certain cognitive, behavioral, or neuropsychiatric symptoms. These models could then be mathematically connected to give rise to an integrative approach towards the multifaceted nature of the disease.

In simpler terms, it is an attempt to computationally assemble many different pieces of the Alzheimer’s puzzle by unifying the enormous body of different works performed over a century of research to a single, inclusive metamodel.

Could computers help predict an individual patient’s future?
If successfully developed and validated, such a framework could eventually help researchers estimate how Alzheimer’s is likely to progress in a particular patient, identify different disease patterns or subtypes, and investigate which biological processes may be most important in that individual.

It could also allow scientists to model possible interventions and compare predicted disease trajectories with and without a particular treatment.

This could have implications for clinical trials as well. Most trials currently compare groups of patients and often target a particular biological process. A more integrated model could potentially help doctors and researchers identify groups of patients who share similar disease mechanisms and determine which treatments are most appropriate to test in each group.

The proposal remains a framework for future research rather than a tool ready for use in the clinic. Building it would require much larger and more diverse datasets, repeated testing, and collaboration across the Alzheimer’s research community. Still, it points toward a significant change in how Alzheimer’s could eventually be approached.