Can AI Machine Learning and Genomics Find Alzheimer’s Drugs?

A new study shows how an AI tool could help with drug repurposing.

Posted Feb 24, 2021

Coffee/Pixabay
Source: Coffee/Pixabay

What if a new treatment for Alzheimer’s disease exists today among existing U.S. Food and Drug Administration (FDA) approved drugs? A new peer-reviewed study published last week in Nature Communications by researchers at Harvard Medical School and Massachusetts General Hospital shows how an AI machine learning framework combined with genomics can help predict drug repurposing candidates for Alzheimer’s disease.

There are an estimated 50 million people living with Alzheimer’s disease, a neurodegenerative disorder, and other forms of dementia globally according to the World Alzheimer Report 2018. In the United States, 5.8 million people are affected by Alzheimer’s disease—two-thirds of whom are women.

This devastating disease without a cure impacts not only the patient, but also families and caregivers. There are over 16 million people in the U.S. caring for those with Alzheimer’s according to an article published today in Time by Maria Shriver, founder of the Women’s Alzheimer’s Movement, and George Vradenburg, co-founder of UsAgainstAlzheimer’s.

There are five FDA-approved medications and no cure for Alzheimer’s disease, according to the Alzheimer’s Association 2019 report. Cholinesterase inhibitors donepezil (2010 FDA approval), galantamine (2001 FDA approval), and rivastigmine (2000 FDA approval) work by increasing the amount of the chemical messenger acetylcholine by blocking acetylcholinesterase that breaks it down according to the same report. Memantime (2003 FDA approval) is an NMDA receptor antagonist that reduces abnormal brain activity, according to the U.S. National Library of Medicine’s MedlinePlus. The final approved drug is a combination of donepezil and memantine (2014 FDA approval).

The process of drug discovery and development is typically a long and arduous process that may span over a decade—with no guarantee of gaining approval from the FDA. There are roughly over 20,000 approved drugs, according to the FDA. To comb through and evaluate each of the 20,000 medications for repurposing for Alzheimer’s treatment is a complex and time-consuming undertaking. Can artificial intelligence (AI) machine learning accelerate the search for drug repurposing for Alzheimer’s disease?

“The approach we developed, drug repurposing in Alzheimer’s disease (DRIAD), involves a machine learning framework that quantifies the association between the stage of AD (early, mid, or late) as defined by Braak staging and any biological process or response that can be characterized by a list of gene names,” wrote researchers Steve Rodriguez, Artem Sokolov, Clemens Hug, Peter Todorov, Nienke Moret, Sarah Boswell, Kyle Evans, George Zhou, Nathan Johnson, Bradley Hyman, Peter Sorger, and Mark Albers.

“We used DRIAD to look for associations between the pathological stage of AD and genes that are differentially expressed when a small molecule drug is applied to a culture of terminally differentiated neuronal pro-genitor cells, which comprises a mix of neurons, astrocytes, and oligodendrocytes,” the researchers wrote.

In order for the machine learning framework to identify associations between gene lists and Alzheimer’s disease, the team used input data from mRNA expression profiles from human brains at various stages of Alzheimer’s and a dataset of drug-associated gene lists. The machine learning framework trains and evaluates a predictor that can spot disease categories from mRNA expression levels.

“DRIAD is applied to lists of genes arising from perturbations in differentiated human neural cell cultures by 80 FDA-approved and clinically tested drugs, producing a ranked list of possible repurposing candidates,” reported the scientists. “Top-scoring drugs are inspected for common trends among their targets. We propose that the DRIAD method can be used to nominate drugs that, after additional validation and identification of relevant pharmacodynamic biomarker(s), could be readily evaluated in a clinical trial.”

Copyright © 2021 Cami Rosso. All rights reserved.