Tech Explained: Here’s a simplified explanation of the latest technology update around Tech Explained: How AI Is Reshaping Alzheimer’s Disease Research in Simple Termsand what it means for users..
AI has been promised as a game-changer for Alzheimer’s disease research; however, translating technical advances into real-world impact can be a challenge.
As the volume and complexity of Alzheimer’s data have exploded, a new generation of AI tools is starting to reveal patterns that were previously invisible: biological signals that emerge years before symptoms, digital markers that redefine diagnosis, and discovery pipelines that operate at unprecedented scale.
A recent special collection in The Journal of Prevention of Alzheimer’s Disease (JPAD), commissioned by Gates Ventures and the Alzheimer’s Disease Data Initiative (AD Data Initiative), explores how AI is being applied across diagnosis, data integration, drug discovery, and clinical trials, alongside the risks of bias and uneven translation. The special issue features two opening editorials and nine in-depth essays from leading scientists.
Technology Networks spoke with Dr. Niranjan Bose, the interim executive director of the AD Data Initiative and managing director (Health & Life Sciences Strategy) at Gates Ventures LLC, where he serves as the science advisor to Bill Gates.
Bose discussed where AI is already delivering value, what has changed in the past two years to move the field beyond hype, and what success for AI in Alzheimer’s research should ultimately look like for patients and families.
Looking across this special collection, where do you think AI is furthest along in Alzheimer’s research?
Niranjan Bose, PhD (NB):
We’ve made great progress in leveraging AI to identify markers for early Alzheimer’s diagnosis. One big theme explored in the JPAD special issue is how AI can identify subtle patterns in imaging, fluid biomarkers, and even speech, that can enable diagnosis years before symptoms become apparent.
These learnings are already being used to inform the development of new diagnostic tools. For example, in the issue, Dr. Liming Wang and colleagues demonstrate how AI-enabled speech analysis can identify pre-symptomatic cognitive decline with remarkable sensitivity.
Dr. Rhoda Au and colleagues make a compelling case for reinventing the ‘N’ in the A/T/N diagnostic framework to include digital and AI-derived biomarkers that could transform our diagnostic toolkit in ways that would have been unimaginable just a few years ago.
The A/T/N diagnostic framework
The A/T/N diagnostic framework is a biomarker-based system for classifying Alzheimer’s disease that organizes evidence into three categories: A for amyloid pathology, T for tau pathology, and N for neurodegeneration or neuronal injury.
What makes large-scale proteomics especially valuable for AI-driven discovery compared with genomics or imaging alone?
Large amounts of quality data are required to train foundational AI models for disease-specific research. Large-scale proteomics is especially valuable because the proteome reflects what is happening in cells and tissues due to the surrounding environment in real-time. Proteins are also the direct targets of most drugs. Multi-modal datasets, including a combination of genomics, proteomics, imaging, and clinical data, should help accelerate model development efforts.
We need more massive, harmonized datasets like the Global Neurodegeneration Proteomics Consortium’s (GNPC) V1 Harmonized Data Set, which has over 250,000 protein measurements and counting, to identify potential biomarkers and targets for drug development. These types of datasets can only be built when you have a coalition of the willing to contribute expertise, data, and resources, and technical infrastructure like the AD Workbench, which provides a secure, cloud-based platform to collaborate.
What changed in the last 12–24 months that makes AI in Alzheimer’s research meaningfully different today than the previous wave of hype?
AI capabilities themselves have evolved dramatically in the past year. We’ve moved from generative tools to advanced “agentic” systems that can reason, plan, and learn autonomously. In August 2025, the AD Data Initiative launched a new prize to solicit the best ideas on how to use agentic AI for Alzheimer’s research, and we’re excited by the proposals we received. Five finalists will be making their pitches in March at the Alzheimer’s Disease and Parkinson’s Disease conference in Copenhagen.
We also now have the data infrastructure AI needs—large, harmonized datasets with high-quality data built through efforts like the GNPC—to be able to train models to generate meaningful insights for Alzheimer’s disease.
Put together, these developments put us in the right time frame to leverage AI meaningfully for Alzheimer’s research.
Where do you see the biggest risk of AI failing to translate into real benefits for patients?
The biggest risk is that AI fails to benefit those who need it most. If models are trained mainly on data from well-resourced, homogeneous populations, they may be inaccurate for underrepresented groups that often face higher Alzheimer’s risk.
In the issue, Dr. Vijaya Kolachalama, Vijay Sureshkumar, and Au describe how skewed imaging datasets could drive underdiagnosis in minority populations, while models based on biomarkers from homogenous datasets may misestimate risk when genetic factors like APOE4 vary by ethnicity.
Dr. Andrew E. Welchman and Dr. Zoe Kourtzi further note that current Alzheimer’s trials over-represent highly educated, research-engaged participants, with minorities still making up only a small fraction, despite higher prevalence.
Without intentional focus on inclusive data, equitable model development, and validation across diverse settings, AI could widen existing disparities instead of closing them—a scientific and ethical failure.
One group aware of this gap is the GNPC—they’re actively working to bring in cohorts from diverse populations across Latin America, Africa, and Asia in the next iteration of the GNPC’s harmonized dataset.
What does success for AI in Alzheimer’s research look like?
With 55 million people worldwide currently living with Alzheimer’s and related dementias, a number that could triple by 2050 without new interventions, the stakes couldn’t be higher.
The hope is that AI helps bend that curve and change outcomes in ways patients and families can feel. That means earlier, more accurate diagnostics that are available in every primary care provider’s office, faster development of effective therapies thanks to AI-identified targets and optimized (including possibly shorter) trial designs, and more personalized treatment strategies that match the right intervention to the right person at the right time.
If AI disappeared tomorrow, what progress would we immediately lose?
If AI disappeared tomorrow, we’d lose our ability to work at the scale, speed, and complexity this disease demands.
The sheer volume of literature and multimodal data on dementia being produced every day is impossible for one human researcher to process. Without AI-enabled tools to support researchers, discovery would continue at the same or a slower rate than it is now, and key insights could remain buried in information overload.