How AI, Graphs, and GraphRAG Are Set to Revolutionize Alzheimer's Gene Research and Treatment

An estimated 6.9 million Americans aged 65 and older are living with Alzheimer’s. Without major medical breakthroughs, that number is projected to rise to 13.8 million in the U.S. by 2060, and to 139 million worldwide by 2050. Humanity is putting enormous effort into preventing those numbers from ever becoming reality, but so far progress has been slow, and the challenge remains daunting.

We may now have a powerful new ally in the fight against Alzheimer’s disease: AI. However, the real story isn't just about AI achieving some kind of magical breakthrough. It's about how healthcare providers and medical researchers are combining AI with advanced software tools and database technologies—particularly knowledge graphs and GraphRAG (Graph Retrieval-Augmented Generation).

These technologies are poised to revolutionize the way AI delivers accurate, contextually relevant insights, enhancing its potential far beyond what AI alone can achieve.

Helping spot the hidden patterns that otherwise slip through

Leading this effort is Cedars-Sinai Medical Center, one of the largest nonprofit academic medical centers in the U.S. Researchers at Cedars-Sinai are leveraging graph technology to drive what they refer to as “Knowledge-Aware” automated machine learning, focusing on two key areas of Alzheimer’s disease research: uncovering potential genetic clues that have previously gone unnoticed, and identifying existing drugs within our pharmacopeia that could be repurposed to enhance the arsenal of treatments available for Alzheimer's.

The goal is not research for the sake of research, but to drive innovation in treatment. And first fruit of the Center’s focus on this is AlzKB, a publicly accessible Alzheimer's disease knowledge base. AlzKB is being utilized to guide machine learning systems toward insights that might otherwise go undetected, or would take human researchers decades to uncover independently.

Jason Moore, who leads the AlzKB effort within Cedars-Sinai’s Department of Computational Biomedicine, explains: “Our goal is to inform machine learning so that it can do a better job of things like feature selection, model selection, and model interpretation.”

To get there, AlzKB is built on more than 20 distinct knowledge sources, combining insights on genes, known genetic links to Alzheimer’s, relationships between genes and drugs, as well as broader biological context such as biochemical pathways, symptoms, and more.

The current version contains 234,000 nodes and 1.67 million “edges” (relationships) between concepts, all structured around a detailed Alzheimer’s ontology (data dictionary) which helps the system understand how different entities connect, which drugs may treat Alzheimer’s, how one gene regulates another, and how these influences cascade.

What’s especially exciting is how the team has engineered AlzKB to be widely usable—by deliberately integrating advanced techniques like large language models (LLMs) and retrieval-augmented generation (RAG). This means researchers and clinicians don’t need to speak a specialist language like Cypher to extract insights; they can simply ask questions in plain English.

That’s made possible by a custom interpretation layer that acts as a bridge between the core LLM and the complex data inside the knowledge graph. As Jason Moore puts it: “We want our biology and clinical users to be able to query the knowledge base without having to learn to program to query the knowledge graph.”

Unaided by graph technology, an LLM is no more accurate than flipping a coin

First port of call for doing this was, of course, publicly available LLMs—but Moore had a pretty good sense of what would happen if he tried a standard, out-of-the-box approach.

“That led us to look at ChatGPT, but we didn't expect ChatGPT to work well on this particular problem,” he states.

Why? Because an off-the-shelf LLM simply can’t understand what’s inside the knowledge graph the team had built. A knowledge graph is a higher-order synthesis of countless pieces of data—connections, context, and relationships. That context does not exist in the public domain in a form that’s searchable or structured enough for any LLM to have been trained on.

“ChatGPT can answer a question about a gene or a drug or Alzheimer's, but it cannot really put all those entities together and understand their relationships in any kind of a complete way.”

The real deal-breaker, though, was that all these limitations added up to something serious: for critical Alzheimer’s research or treatment questions, the system just wasn’t reliable. As Moore says, “ChatGPT was no better right out of the box than flipping a coin at answering questions.”

A better graph-based approach

However, by integrating the power of knowledge graphs with GraphRAG—and combining that with a recent innovation from Hugging Face known as Graph of Thoughts—Cedars-Sinai has radically improved the effectiveness of its natural language interface, unlocking far more of the insight and intelligence AlzKB was built to provide.

“Graph gives us a higher-level synthesis of the knowledge than we would get in any other way,” he explains. In other words, the knowledge graph doesn’t just store facts, but actually helps the knowledge base reason across them, turning fragments into genuine insights.

But as powerful as it is for researchers to query something like AlzKB, the real breakthrough may lie in the strides Moore and his colleagues at Cedars-Sinai are making in evolving machine learning itself—especially machine learning that’s grounded not just in data, but in a nuanced, biological understanding of Alzheimer’s.

“There have been a lot of large-scale genetic studies of Alzheimer's, and from these very large patient datasets science has so far identified about 100 of the 20,000 genes in the human genome as risk factors for Alzheimer disease,” he says.

Moore says the aim is to surface new genetic risk factors and, ultimately, new drugs for Alzheimer’s. But rather than tread the well-worn path—where global research has already intensely focused on the top 100 genes linked to the disease and a catalogue of known drug leads—he’s deliberately looking elsewhere.

The metaphor used to describe this approach and how AI can support search in novel and hopefully interesting areas is the distinction between the visible and the so far invisible. “Lots of people are looking under the lamppost of a known gene,” he says. “What I’m interested in is the novel discoveries over ‘there,’ out in the dark.”

Using knowledge graph to inform machine learning

In practical terms, that means using the knowledge graph to sharpen how features are selected in the automated machine learning pipelines Cedars-Sinai is developing—and the implications are significant.

Even in these early stages, the research is surfacing previously overlooked genetic contributors to Alzheimer’s, and pointing to a tantalizing possibility: that common, everyday medications—like Temazepam, typically used to manage insomnia (a frequent symptom of Alzheimer’s), and even Ibuprofen, prescribed for headaches—might be repurposed or re-engineered to treat the disease at its root.

For Moore, the significance of this combination of graph, advanced data analysis and AI itself has big implications. “I think we've shown how we can tailor large language models and accurately query a big Alzheimer’s database,” he says, “plus demonstrated how we can use the knowledge in the knowledge graph to inform machine learning to give us new ideas for treating this disease.”

Can graphs get us closer to AI-personalized, optimized medical R&D?

The next frontier is a fully automated machine learning system—again powered by the knowledge graph. Soon, AlzKB users will be able to type prompts like, “Show me the genes tied to this drug and this disease,” or “Create a dataset with only these specific genes,” and let the AI take over. From there, it will run algorithms, highlight the most important features, and provide a clear, interpretable output that a scientist can immediately act on—accelerating real progress in Alzheimer’s research today, not someday.

What’s clear—here and in many other cutting-edge projects—is that developers are increasingly turning to graphs and GraphRAG to tap into the full power of big medical data and AI. This isn’t a vision of the future. It’s happening right now.

Topics: AI & Digital