Using artificial intelligence (AI), researchers at the University of California, San Francisco, have been able to forecast an individual’s likelihood of developing Alzheimer’s up to seven years before any symptoms appear. Researchers found early risk factors that affected both men and women, as well as a few that were gender-specific, such as osteoporosis in women and erectile dysfunction and an enlarged prostate in men.
Alzheimer’s is the most common form of dementia, primarily affecting individuals over the age of 65. Its characteristic features include progressive memory loss, cognitive decline, and a range of neurological abnormalities, such as the buildup of tau tangles and amyloid-beta plaques in the brain. These abnormal modifications disturb regular brain cell activity, resulting in illness symptoms and, eventually, severe disability.
Despite continuous research, Alzheimer’s currently has no known cure; instead, current therapies primarily aim to manage symptoms rather than slow or reverse the illness’s course.
Early Alzheimer’s detection provides a critical advantage: the possibility of earlier intervention, which could drastically modify the illness’s course or reduce its effects. Conventional techniques to diagnose Alzheimer’s disease, such as cognitive evaluations and biomarker analysis, are only frequently used after the illness has developed, sometimes too late to allow for the best course of treatment.
In one such attempt to find a way to diagnose Alzheimer’s earlier, researchers at UC San Francisco have created a machine learning model that can predict the disease up to seven years before any symptoms appear.
While researchers discovered several early risk factors common to both men and women, they also identified a few that were gender-specific, such as erectile dysfunction and an enlarged prostate in males and osteoporosis in women.
About the Recent Research
The study, published in Nature Aging on February 21, 2024, highlights how artificial intelligence (AI) has the potential to significantly advance our understanding of and ability to diagnose complicated disorders like Alzheimer’s early [1].
The research team used the vast electronic health databases at the UCSF Medical Center to build their predictive models. Based on expert-level clinical diagnoses, researchers were able to identify 749 individuals from this pool as having Alzheimer’s disease and 250,545 controls who did not have a dementia diagnosis.
Using Random Forest (RF) models—a type of machine learning algorithm—was key to the process because it can handle the intricate, non-linear correlations frequently found in medical data. The researchers trained the models with a wide variety of clinical data points taken from electronic health records, such as demographics, medical problems, drug exposures, and abnormal laboratory measures.
The results showed that the machine learning models could reliably forecast the development of Alzheimer’s disease up to seven years ahead of time with high accuracy (72%). Alongside clinical data, the addition of visit-related and demographic information improved the predictive accuracy of the models further.
Alzheimer’s predictors in men and women
Besides the early risk factors identified for both men and women, researchers uncovered a few gender-specific risk variables.
Several variables, including hypertension, high cholesterol, and vitamin D insufficiency, have emerged as the leading predictors of Alzheimer’s in men and women.
Erectile dysfunction and an enlarged prostate were also predictors in men. The study identified osteoporosis as another critical predictor for women, pointing to a potential gender-specific mechanism or predisposition to the condition.
However, not all women who have osteoporosis will eventually get Alzheimer’s. The combination of disorders enabled the model to predict Alzheimer’s onset. The discovery that osteoporosis is one prognostic factor for females underscores the biological interaction between bone health and the risk of dementia, according to the researchers.
Delving deeper into the biological mechanisms
To further explore the biological mechanisms underlying the prediction capabilities of their model, the researchers used SPOKE (Scalable Precision Medicine Oriented Knowledge Engine), a potent tool created at UCSF, in conjunction with public molecular databases.
SPOKE is a “database of databases” created in the lab of Sergio Baranzini, a neurology professor and a UCSF Weill Institute for Neurosciences member. Researchers may go through enormous volumes of data via this cutting-edge technique to find patterns and possible biological targets for therapeutic intervention.
SPOKE established that the APOE4 variant of the apolipoprotein E gene is responsible for the association between high cholesterol and Alzheimer’s disease. This relationship is well-known among scientists. However, combining SPOKE with genetic databases led to a discovery: a relationship between osteoporosis and Alzheimer’s, particularly in women.
A variant in the MS4A6A gene—which is not as well-known in the context of Alzheimer’s research—led to the discovery of this connection. The identification of this correlation highlights the potency of merging sophisticated computational tools such as SPOKE with copious genetic data, enabling focused investigation into the molecular mechanisms implicated in Alzheimer’s disease and perhaps steering the creation of novel therapeutic approaches.
Limitations of the study
The results signify a significant breakthrough in the battle against Alzheimer’s. Notwithstanding the encouraging outcomes, the research is subject to several limitations, such as the difficulties in analyzing data from electronic health records, the possibility of biases in cohort selection, and the requirement for ongoing model retraining to accommodate evolving clinical procedures. The study’s predictive models must be verified in broader and more varied groups to confirm their accuracy and generalizability.
The researchers have high hopes for the applicability of their techniques to other difficult-to-diagnose conditions like endometriosis and lupus.
References
- Tang, A.S., Rankin, K.P., Cerono, G., Miramontes, S., Mills, H., Roger, J., Zeng, B., Nelson, C., Soman, K., Woldemariam, S. and Li, Y., 2024. Leveraging electronic health records and knowledge networks for Alzheimer’s disease prediction and sex-specific biological insights. Nature Aging, pp.1-17.
- AI can predict Alzheimer’s disease up to seven years before symptoms appear, study finds. PsyPost. https://www.psypost.org/ai-can-predict-alzheimers-disease-up-to-seven-years-before-symptoms-appear-study-finds/. Published Online: 24th February, 2024. Accessed: 8th March, 2024.
- AI finds several early risk factors to predict Alzheimer’s 7 years early. Medical News Today. https://www.medicalnewstoday.com/articles/ai-finds-several-early-risk-factors-predict-alzheimers-7-years-early. Published Online: 29th February, 2024. Accessed: 8th March, 2024,
- AI Predicts Alzheimer’s 7 Years Early. Neuroscience News.com. https://neurosciencenews.com/ai-alzheimers-25642/. Published Online: 21st February, 2024. Accessed: 8th March, 2024.