, raising the prospect of early intervention when effective treatments become available.
In a study, investigators used six AI methodologies, including Deep Learning, to assess blood leukocyte epigenomic biomarkers. They found more than 150 genetic differences among study participants with Alzheimer’s disease in comparison with participants who did not have Alzheimer’s disease.
All of the AI platforms were effective in predicting Alzheimer’s disease. Deep Learning’s assessment of intragenic cytosine-phosphate-guanines (CpGs) had sensitivity and specificity rates of 97%.
“It’s almost as if the leukocytes have become a newspaper to tell us, ‘This is what is going on in the brain,’ “ lead author Ray Bahado-Singh, MD, chair of the department of obstetrics and gynecology, Oakland University, Auburn Hills, Mich., said in a news release.
The researchers noted that the findings, if replicated in future studies, may help in providing Alzheimer’s disease diagnoses “much earlier” in the disease process. “The holy grail is to identify patients in the preclinical stage so effective early interventions, including new medications, can be studied and ultimately used,” Dr. Bahado-Singh said.
“This certainly isn’t the final step in Alzheimer’s research, but I think this represents a significant change in direction,” he told attendees at a press briefing.
The findings were published online March 31 in PLOS ONE.
Silver tsunami
The investigators noted that Alzheimer’s disease is often diagnosed when the disease is in its later stages, after irreversible brain damage has occurred. “There is currently no cure for the disease, and the treatment is limited to drugs that attempt to treat symptoms and have little effect on the disease’s progression,” they noted.
Coinvestigator Khaled Imam, MD, director of geriatric medicine for Beaumont Health in Michigan, pointed out that although MRI and lumbar puncture can identify Alzheimer’s disease early on, the processes are expensive and/or invasive.
“Having biomarkers in the blood ... and being able to identify [Alzheimer’s disease] years before symptoms start, hopefully we’d be able to intervene early on in the process of the disease,” Dr. Imam said.
It is estimated that the number of Americans aged 85 and older will triple by 2050. This impending “silver tsunami,” which will come with a commensurate increase in Alzheimer’s disease cases, makes it even more important to be able to diagnose the disease early on, he noted.
The study included 24 individuals with late-onset Alzheimer’s disease (70.8% women; mean age, 83 years); 24 were deemed to be “cognitively healthy” (66.7% women; mean age, 80 years). About 500 ng of genomic DNA was extracted from whole-blood samples from each participant.
The researchers used the Infinium MethylationEPIC BeadChip array, and the samples were then examined for markers of methylation that would “indicate the disease process has started,” they noted.
In addition to Deep Learning, the five other AI platforms were the Support Vector Machine, Generalized Linear Model, Prediction Analysis for Microarrays, Random Forest, and Linear Discriminant Analysis.
These platforms were used to assess leukocyte genome changes. To predict Alzheimer’s disease, the researchers also used Ingenuity Pathway Analysis.