Artificial intelligence (AI) is a broad field that includes many different subtypes, ranging from apps capable of writing poetry to algorithms that detect patterns easily missed by humans. Recently, AI modeling has played a significant role in advancing an Alzheimer's disease study. Researchers at the University of California, San Diego (UC San Diego) utilized AI to uncover that a gene commonly identified as a marker of Alzheimer's might also be a causative factor. This discovery underscores a major challenge in Alzheimer's research: distinguishing the changes caused by the disease from the changes that actually trigger it. The focus of this study was an enzyme called phosphoglycerate dehydrogenase (PHGDH) and the gene encoding it. Prior research by the team had shown that this gene tends to be more active in individuals with rapidly progressing Alzheimer's. However, what was unclear was the mechanism behind this connection. Using AI, the researchers modeled the three-dimensional structure of the PHGDH enzyme in greater detail, revealing a previously unknown function: it appears to switch other specific genes on and off. Further investigations demonstrated that PHGDH interacts with two genes inside astrocytes—brain cells involved in supporting neurons—in ways that disrupt the brain’s capacity to manage inflammation and clear waste. The researchers believe this interaction may represent a critical tipping point that initiates Alzheimer's, explaining the link between PHGDH and the disease. “This discovery really required state-of-the-art AI to precisely determine the enzyme’s 3D structure, ” notes Sheng Zhong, a bioengineer at UC San Diego. Subsequently, the team sought to develop methods to partially inhibit PHGDH.

Ideally, a drug would block PHGDH’s gene-regulating activity in astrocytes, while preserving its essential enzymatic functions. They identified a molecule named NCT-503 that meets these criteria. AI modeling was again employed to analyze the structure of NCT-503 and its interaction with PHGDH. It appears that NCT-503 binds to a specific pocket in PHGDH, preventing its unauthorized gene-switching actions. Although it will take substantial time before a drug for Alzheimer's can be developed based on this finding, initial research has shown that treatments derived from NCT-503 can effectively regulate PHGDH activity in mouse models of the disease. Mice treated with the molecule demonstrated improved performance in memory and anxiety-related tests. “Now there is a therapeutic candidate with demonstrated efficacy that holds promise for further clinical development, ” says Zhong. “There could be entirely new classes of small molecules that can be harnessed for future therapeutics. ” Importantly, NCT-503 is capable of crossing the blood-brain barrier to reach neurons and their associated cells, enhancing the potential impact of this research. Drugs based on NCT-503 could also be administered orally. While unraveling the complexity of Alzheimer’s disease—including the various contributing factors from environmental challenges to inherited genetics—remains a slow process, each new study brings us closer to effective treatments and better management of the condition. “Unfortunately, treatment options for Alzheimer’s disease remain very limited, ” Zhong reflects. “At present, treatment responses are far from optimal. ”