Summary: Researchers at UCSF have developed an innovative AI approach to predict the onset of Alzheimer’s Disease up to seven years before symptoms appear. By analyzing patient records with machine learning, the study highlights the importance of high cholesterol and osteoporosis, especially in women, as crucial indicators. Integrating clinical data with genetic databases using tools like UCSF’s SPOKE has enabled the identification of genes associated with Alzheimer’s, offering new possibilities for early detection and understanding the complex relationship between various health conditions and Alzheimer’s risk. This advancement has the potential to enhance precision medicine for Alzheimer’s and other complex diseases.
Key Findings:
1. Early Detection with AI: Machine learning applied to clinical data can forecast Alzheimer’s onset with 72% accuracy up to seven years before symptoms manifest.
2. Identification of Significant Predictors: High cholesterol and osteoporosis are key predictors of Alzheimer’s, with osteoporosis particularly impactful for women.
3. Genetic Revelations: Through UCSF’s SPOKE, researchers have linked Alzheimer’s risk to specific genes, revealing a connection between osteoporosis and Alzheimer’s in women via the MS4A6A gene.
Source: UCSF
UC San Francisco researchers have introduced a method to predict Alzheimer’s Disease up to seven years before symptoms appear by utilizing machine learning to analyze patient records.
The primary conditions influencing Alzheimer’s prediction were high cholesterol and, notably for women, osteoporosis.
This study underscores the potential of artificial intelligence (AI) in recognizing patterns in clinical data, which can then be leveraged to explore extensive genetic databases and uncover underlying risk factors. The ultimate aim is to expedite Alzheimer’s diagnosis and treatment, along with other intricate diseases.
The researchers anticipate extending this methodology to other challenging-to-diagnose diseases such as lupus and endometriosis. Credit: Neuroscience News
“This signifies the initial phase of employing AI on routine clinical data not only to detect risk at the earliest possible stage but also to comprehend the biological mechanisms behind it,” mentioned Alice Tang, the lead author of the study and an MD/PhD student in the Sirota Lab at UCSF.
“The effectiveness of this AI methodology lies in identifying risk based on combinations of diseases.”
The study was published on February 21, 2024, in Nature Aging.
Clinical Data and Predictive Potential
For an extended period, scientists have been striving to reveal the biological drivers and early indicators of Alzheimer’s Disease, a degenerative form of dementia leading to memory loss. Alzheimer’s affects approximately 6.7 million Americans, with nearly two-thirds being women. While the disease’s risk increases with age, and women generally outlive men, this alone does not explain the higher prevalence among women.
By scrutinizing UCSF’s clinical database comprising over 5 million patients, the researchers identified co-occurring conditions in patients diagnosed with Alzheimer’s at UCSF’s Memory and Aging Center, enabling them to predict who would develop the disease up to seven years in advance with 72% accuracy.
Various factors such as hypertension, high cholesterol, and vitamin D deficiency were predictive for both genders. In men, erectile dysfunction and an enlarged prostate were also predictive, while for women, osteoporosis emerged as a particularly crucial predictor.
However, having osteoporosis does not guarantee Alzheimer’s development.
“Our model’s ability to predict AD onset stems from the combination of various diseases,” Tang explained. “The identification of osteoporosis as a predictive factor for females underscores the biological correlation between bone health and dementia risk.”
Advancing Precision Medicine
To delve into the biological underpinnings of the predictive model, the researchers delved into public molecular databases and utilized a specialized UCSF tool called SPOKE (Scalable Precision Medicine Oriented Knowledge Engine), developed by Sergio Baranzini, PhD, a neurology professor and member of the UCSF Weill Institute for Neurosciences.
SPOKE acts as a compendium of databases that researchers can utilize to identify patterns and potential molecular targets for therapeutic interventions. While confirming the well-established link between Alzheimer’s and high cholesterol through a variant of the apolipoprotein E gene, APOE4, SPOKE, in conjunction with genetic databases, also unveiled a connection between osteoporosis and Alzheimer’s in women through a variant in a less-known gene, MS4A6A.
The researchers aim to apply this methodology to other challenging-to-diagnose diseases like lupus and endometriosis.
“This study exemplifies how we can harness patient data and machine learning to predict the likelihood of Alzheimer’s onset in patients and gain insights into the underlying reasons,” stated Marina Sirota, PhD, the senior author of the study and an associate professor at the Bakar Computational Health Sciences Institute at UCSF.
Authors: The study involved additional UCSF co-authors such as Katherine P. Rankin, PhD, Gabriel Cerono, MD, Silvia Miramontes, MIDS, Hunter Mills, MS, Jacquelyn Roger, PhD student, Billy Zeng, MD, Charlotte Nelson, PhD, Karthik Soman, PhD, Sarah Woldemariam, Yaqio Li, PhD, Albert Lee, MADS, Riley Bove, MD, Tomiko Oskotsky, MD, Zachary Miller, MD, Isabel Allen, PhD, Stephan J. Sanders, PhD, and Sergio Baranzini, PhD.
Funding: The primary support for this study was provided by the National Institute on Aging (grant R01AG060393). Additional support came from the Medical Scientist Training Program (T32GM007618) and F30 Fellowship (1F30AG079504-01).
About the Research
Author: Victoria Colliver
Source: UCSF
Contact: Victoria Colliver – UCSF
Image: Image credited to Neuroscience News
Original Research: Open access.
“Leveraging electronic health records and knowledge networks for Alzheimer’s disease prediction and sex-specific biological insights” by Alice Tang et al. Nature Aging
Abstract
Leveraging electronic health records and knowledge networks for Alzheimer’s disease prediction and sex-specific biological insights
The study demonstrates that electronic health records from the University of California, San Francisco, in conjunction with knowledge networks like SPOKE, enable (1) the prediction of Alzheimer’s Disease onset, (2) the prioritization of biological hypotheses, and (3) the contextualization of sex-specific differences.
By training random forest models and forecasting AD onset in a cohort of 749 individuals with AD and 250,545 controls, the study achieved a mean area under the receiver operating characteristic curve ranging from 0.72 (7 years prior) to 0.81 (1 day prior). The researchers also utilized matched cohort models to identify conditions with predictive potential before the onset of Alzheimer’s.
Knowledge networks revealed shared genes between top predictors and AD (e.g., APOE, ACTB, IL6, INS). Genetic colocalization analysis supported the association of AD with hyperlipidemia at the APOE locus, as well as a stronger association of AD in females with osteoporosis near the MS4A6A locus.
This study showcases how clinical data can be harnessed for early prediction of Alzheimer’s Disease and the identification of personalized biological hypotheses.
