AI-powered databases boost the Alzheimer’s drug discovery process

Researchers studying Alzheimer’s disease are using artificial intelligence-powered databases to accelerate the drug discovery process by making it easier to sift through vast amounts of biomedical data.

By using those technologies, scientists at the United Kingdom’s Oxford Drug Discovery Institute can speed up the work of digging through journals and databases by nearly ten times—helping to more quickly prioritize which genes or proteins should be selected for further work to generate potential Alzheimer’s drugs, it said.

Biologists at the Oxford Drug Discovery Institute had selected 54 genes from a genome-wide association study that were related to the immune system, all of which are likely targets for lab testing, said Emma Mead, its chief scientific officer. Those targets can include biological structures like genes or proteins, which potential drugs aim to affect.

Picking Alzheimer’s targets can be particularly tricky because there are so many genes that can increase the risk of developing disease, and because the disease has so many confounding environmental and socioeconomic risk factors, Mead said.

But it took leveraging a knowledge graph, a database technology popularized by Google over a decade ago for its search engine, for staff to more quickly decipher those targets’ properties across a large number of sources—from the U.S. National Library of Medicine’s PubMed to various scientific journals and its own datasets.

Knowledge graphs—which are like databases that represent information similar to maps— can show relationships between people, ideas and documents. In more recent years, they’ve been used by industries like digital retail to provide customized recommendations to online shoppers.

Among businesses, knowledge graphs are being used alongside a method called retrieval-augmented generation, or RAG, to help fine-tune the general-purpose AI models offered by companies like Anthropic or OpenAI. AI models can also be linked up with vector databases, which are a different format of storing data that represent data as “vectors."

For instance, the method of connecting databases with AI helps businesses customize the AI chatbots they build for employees to reference company policies.

Healthcare and life sciences organizations, in particular, can benefit from organizing their disparate data sources into a map-like relationship, rather than traditional relational databases, said Radu Miclaus, an analyst at market research and IT consulting firm Gartner.

By using a knowledge graph, scientists can trace where information on a specific gene or protein was found, including the exact article that cites a particular biological relationship, or the database it comes from, said Martina Markova, a senior product manager at Graphwise, a company that builds knowledge graphs for businesses.

The Oxford Drug Discovery Institute’s biologists worked with Graphwise to customize a large-scale knowledge graph of their life sciences research information. That process helped the institute’s biologists speed up the time it took to evaluate 54 genes from a few weeks to a few days, it said, and is helping them identify biomarkers that may be linked to the genes.

With a subset of the genes it has prioritized, Mead’s team plans to perform more experimental validation. That involves confirming the targets can cause changes in brain cells that contribute to disease, and determining whether the targets are “druggable," she said.

Most importantly, greater availability of biomedical data in recent years has been a boon to researchers—if they’re able to decipher it. That’s why tools like AI and knowledge graphs can provide a much needed boost to scientists and researchers, whether or not they have a bioinformatics background. “Otherwise, you feel a bit lost with the data," Mead said.

Write to Belle Lin at belle.lin@wsj.com