Verified by 
AI Improves Drug Screening for Alzheimer’s Disease
New AI machine learning method can be used for AD and a variety of diseases.
Posted June 2, 2022 Reviewed by Kaja Perina
In a new study, researchers at the University of California, San Francisco (UCSF) apply artificial intelligence (AI) machine learning to find valuable hidden data in archival disease databases that may help accelerate biotechnology research and drug discovery for Alzheimer’s disease (AD) and other conditions.
“Despite extensive drug discovery efforts, no effective treatments exist that prevent or even slow the progression of AD,” the researchers wrote. “Of the many therapeutic targets under investigation, the pathogenic misfolding and accumulation of tau protein into neurofibrillary tangles (NFTs) has emerged as a target mechanism.
Alzheimer’s disease accounts for 60-80 percent of dementia cases and is the most common cause of dementia according to the Alzheimer’s Association. Dementia is the decline in mental ability that interferes with life’s daily activities with symptoms that may impact a person’s ability to remember, think, and reason. An estimated 139 million people worldwide will be living with dementia by 2050 and over 55 million people globally live with dementia in 2020 according to Alzheimer’s Disease International. According to the Women’s Alzheimer’s Movement (WAM), an organization founded by Maria Shriver, an estimated 6 million Americans live with Alzheimer’s disease, of which two-thirds are women and scientists do not know why.
This study applied machine learning to an archival High Content Screening database that had information on the phenotypic effects of small molecules for Alzheimer’s disease treatment.
The researchers sought to improve the extraction of biological insights from imaging data gathered from High Content Screening (HCS), a method that is vital for the drug discovery process.
Their hypothesis was that AI machine learning could find actionable data within the HCS that could provide valuable insights into the biochemical characteristics of an organism which in turn can accelerate biotechnology research and drug discovery.
In microbiology, High Content Screening (HCS), also known as an automated microscope-based screening, is used for research and toxicity screening for drug discovery. It was developed in the mid-1990s and is often used in systems biology research and to discover whether candidate drugs change the course of the disease by enabling scientists to measure and understand the functions of proteins, genes, RNA, and other components of living cells.
High Content Screening is mostly a fluorescence microscopy imaging technique for living cells and organisms where light emitting fluorescent materials are examined under a microscope. Typically, the living cell of the specimen is stained with a fluorescence stain that light up with exposed to short wavelength light such as blue or ultraviolet (UV) light and seen via filters that remove undesired light wavelengths.
The current standard methodology to gain more information from High Content Screening data is to introduce additional biological markers. The drawback is that it may be costly, time-consuming, or overly tedious to add more biological markers for tracking. Moreover, this approach will not work for large, archive datasets where the research has already terminated.
The researcher reported that using their AI machine learning method, they were able to identify new compounds that effectively blocked tau aggregation that were previously not found using existing screening approaches without artificial intelligence.
“We overcame these marker limitations computationally by directly learning the phenotypic relationships between a highly-informative but cumbersome marker and other similar yet more easily-accessible markers,” the researchers wrote. “These hidden relationships were then projected into de novo images that displayed the desired fluorescent signal of the cumbersome marker.”
Additionally, the machine learning algorithm can be used for other diseases, not just Alzheimer’s disease and can be used to look for hidden information in other archival fluorescence microscopy image datasets. The scientists evaluated the generalizability of the AI algorithm on a cancer-based dataset, a different biological environment. Specifically, they applied the machine learning to a functional genomics screen in a common type of bone cancer, osteosarcoma, line.
Through the combination of AI machine learning, fluorescence microscopy, and life sciences databases, disease researchers now have a powerful tool to help accelerate drug discovery and the development of novel therapeutics in the future.
Copyright © 2022 Cami Rosso
