New biomarker tracks cognitive decline in Alzheimer’s disease

Researchers have been working to develop tests to help detect and track dementia early in the disease process. Westend61 on Offset / Shutterstock

In people with Alzheimer’s disease (AD), changes in the brain gradually erode the ability to think and remember. This cognitive decline involves an abnormal buildup of the proteins amyloid beta (Aβ) and tau. Measures of these protein biomarkers through brain scans or tests of cerebrospinal fluid (CSF) in the brain and spinal cord have improved AD diagnosis.

However, some people with high levels of Aβ and tau have no detectable cognitive problems. Existing biomarkers also can't fully account for the speed of progression from mild cognitive impairment to severe dementia, which can take from 2 to 20 years.

To learn more about the factors that affect cognitive decline, an NIH-funded team led by Drs. Hamilton Se-Hwee Oh and Tony Wyss-Coray at Stanford University analyzed CSF samples from about 3,400 people. The samples were from research studies in the U.S., Sweden, and Finland of people both with and without a diagnosis of AD who had volunteered to participate in the studies over many years.

The researchers used large-scale protein analysis, or proteomics, to measure levels of more than 7,000 proteins in each of the CSF samples. They searched for new proteins that might help explain differences in cognitive impairments, or thinking ability, among people with AD. To do so, they integrated their protein data with other data collected in the studies. Those included Aβ and tau measurements from CSF and brain scans, along with measures of cognitive ability, age, sex, and AD risk genes, including APOE. The results appeared in Nature Medicine on March 31, 2025.

The team found hundreds of proteins whose levels correlated with cognitive function. The most significant ones were related to synapse function. Synapses are the connections between neurons. In addition to the buildup of Aβ and tau, the loss of connections between neurons in the brain is a key feature of AD. Two synapse-related proteins, YWHAG and NPTX2, were the most closely related to measures of cognitive impairment.

The researchers used machine learning to search for patterns in the protein data that could reliably predict cognitive impairment. This analysis showed that a ratio of YWHAG:NPTX2 reflected a person’s cognitive impairment better than existing biomarkers for Aβ and tau.

YWHAG goes up in people with memory problems, while NPTX2 goes down. As a result, the YWHAG:NPTX2 ratio increases in people experiencing cognitive decline. It also rises in those at higher risk of advancing to full-blown dementia. These findings suggest that the YWHAG:NPTX2 ratio might be used to help predict the onset of AD symptoms and track disease progression. The researchers also found that this ratio rises somewhat as people age normally.

The team next used machine learning to try to develop a similar biomarker using less invasive proteomic blood tests. They were able to develop a set of protein measurements that correlated with the CSF YWHAG:NPTX2 ratio and could also help predict cognitive decline.

“More study is needed to understand the connection between these synaptic proteins and cognitive decline,” Wyss-Coray explains. “But our findings highlight the weakening and loss of neural connections as a driver of the decline.”

Further work will be needed to develop effective tests for use in the clinic. Such tests might one day be used to help detect memory loss sooner, perhaps even before it begins, to allow for early interventions. They could also help to select people for participation in clinical trials of promising new AD treatments and to measure treatment responses.

—by Kendall K. Morgan, Ph.D.