Reviewed by Lexie CornerApr 28 2025
Researchers at the University of California, San Diego, used artificial intelligence to help investigate a long-standing mystery in Alzheimer's disease and identify a potential therapeutic approach that targets the gene's secondary function. The study was published in the journal Cell.
The study found that a gene, recently identified as a biomarker for Alzheimer's disease, is a direct cause of the condition due to an unrecognized secondary function.
Alzheimer's disease, the most common cause of dementia, affects approximately one in nine individuals aged 65 and older. While mutations in specific genes can lead to Alzheimer's, these genetic links account for only a small portion of all cases.
The majority of patients do not have a mutation in a known disease-causing gene; instead, they develop "spontaneous" Alzheimer's, and the underlying causes of this form remain unclear.
Identifying these causes could lead to significant advancements in the medical care of Alzheimer's patients.
Unfortunately, treatment options for Alzheimer’s disease are very limited. And treatment responses are not outstanding at this moment.
Sheng Zhong, Study Senior Author and Professor, Shu Chien-Gene Lay Department of Bioengineering, UC San Diego Jacobs School of Engineering
Zhong and his team conducted a more detailed investigation into phosphoglycerate dehydrogenase (PHGDH), a gene they had previously identified as a potential blood biomarker for the early detection of Alzheimer's disease.
In a subsequent study, they found that the expression levels of the PHGDH gene were directly correlated with changes observed in the brains of individuals with Alzheimer's disease. Specifically, higher levels of protein and RNA produced by the PHGDH gene were associated with more advanced stages of the disease. According to Zhong, this correlation has been confirmed across multiple patient groups from various medical centers.
Driven by this reproducible correlation, the research team in their most recent study sought to explore whether a causal relationship existed. Using mice and human brain organoids, the researchers found that manipulating PHGDH expression levels had significant effects on Alzheimer's disease: lower levels were linked to reduced disease progression, while higher levels resulted in greater disease advancement. This led the researchers to conclude that PHGDH is indeed a causal gene in spontaneous Alzheimer's disease.
Additionally, with the help of artificial intelligence, the researchers discovered that PHGDH has a previously unrecognized role in initiating a pathway that disrupts the regulation of gene expression in brain cells. This disruption can contribute to the development of Alzheimer's disease.
Moonlighting Role
PHGDH produces an enzyme essential for synthesizing serine, an amino acid that also functions as a neurotransmitter. Since PHGDH's enzymatic activity was its only known function, the researchers initially hypothesized that its metabolic role was linked to the development of Alzheimer's. However, their experiments designed to test this connection produced negative results.
At that time, our study hit a wall, and we didn’t have a clue of what mechanism it is.
Sheng Zhong, Study Senior Author and Professor, Shu Chien-Gene Lay Department of Bioengineering, UC San Diego Jacobs School of Engineering
However, a separate Alzheimer's research project in his laboratory, initially unrelated to PHGDH, led to a critical discovery. A year earlier, that project had identified a key feature of Alzheimer's disease: a widespread disruption in the brain's process of regulating gene activation and deactivation, which is essential for cells to perform their specific functions.
This led the researchers to consider whether PHGDH might have an undiscovered regulatory role in the gene expression control process, prompting them to use advanced AI for further investigation.
Using AI, the team was able to visualize the three-dimensional structure of the PHGDH protein. Within this structure, they identified a substructure that resembled a known DNA-binding domain found in transcription factors. This similarity was evident in the three-dimensional arrangement of atoms, rather than in the linear sequence of amino acids that make up the protein.
Zhong remarked, “It really demanded modern AI to formulate the three-dimensional structure very precisely to make this discovery.”
After identifying this structural similarity, the team demonstrated that this specific substructure enables the PHGDH protein to activate two critical target genes. This activation disrupts the balance of gene expression, triggering a cascade of issues that contribute to the early stages of Alzheimer's disease.
In effect, PHGDH has a previously unrecognized function, distinct from its enzymatic activity, that contributes to the development of spontaneous Alzheimer's disease through a novel pathway.
The study ties back to earlier research: the PHGDH gene produced higher levels of protein in the brains of Alzheimer's patients compared to control brains, and these elevated protein levels in the brain triggered the observed imbalance in gene regulation. While the PHGDH gene is present in everyone, the key difference is in the expression level, or the amount of protein it produces.
Treatment Option
After elucidating the underlying mechanism, the researchers sought to identify a potential therapeutic intervention targeting the disease.
While many current Alzheimer's treatments focus on addressing the abnormal accumulation of beta-amyloid protein in the brain, some research suggests that targeting these plaques may be ineffective, implying that treatment at that advanced stage could be too late. However, the critical pathway identified in this study occurs earlier in the disease process, suggesting that intervening at this stage could prevent amyloid plaque formation from the outset.
Given PHGDH's essential role as an enzyme, previous studies have explored potential inhibitors. One small molecule, NCT-503, particularly interested the researchers because it does not significantly block PHGDH's enzymatic activity (serine production), which they aimed to preserve. Additionally, NCT-503 has the ability to cross the blood-brain barrier, a desirable trait for a potential therapeutic.
The researchers used AI for three-dimensional visualization and modeling, which revealed that NCT-503 can interact with the DNA-binding substructure of PHGDH by binding to a specific pocket. Further experimentation confirmed that NCT-503 does inhibit PHGDH's regulatory function.
When tested in two mouse models of Alzheimer's disease, NCT-503 resulted in a significant reduction in disease progression. The treated mice showed notable improvements in memory and anxiety tests, which were selected because Alzheimer's patients commonly experience cognitive decline and increased anxiety.
The researchers acknowledge certain limitations of their study, including the absence of a perfect animal model for spontaneous Alzheimer's disease. As a result, they were only able to test NCT-503 in mouse models carrying mutations in known disease-causing genes.
Despite these limitations, Zhong believes the results are promising.
Now there is a therapeutic candidate with demonstrated efficacy that has the potential of being further developed into clinical tests. There may be entirely new classes of small molecules that can potentially be leveraged for development into future therapeutics.
Sheng Zhong, Study Senior Author and Professor, Shu Chien-Gene Lay Department of Bioengineering, UC San Diego Jacobs School of Engineering
Zhong further noted that a benefit of small molecules like NCT-503 is their potential for oral administration, in contrast to current Alzheimer's treatments that require intravenous infusions.
The next steps will involve optimizing the compound and conducting the necessary studies to obtain an Investigational New Drug (IND) application from the FDA.
Journal Reference:
Chen, J., et al. (2025). Transcriptional regulation by PHGDH drives amyloid pathology in Alzheimer’s disease. ScienceDirect. doi.org/10.1016/j.cell.2025.03.045