Potential treatments for amyotrophic lateral sclerosis (ALS) and other neurodegenerative diseases may already be there in the form of drugs prescribed for other conditions. A team of researchers from Stanford University Lawrence Livermore National Laboratory (LLNL) and University of California, Los Angeles (UCLA) are trying to find them using artificial intelligence and machine learning (AI/ML).
Reusing existing drugs is one of the best ways to quickly provide treatment, as new drug clinical trials can take up to 5-7 years. AI/ml can be even faster. By analyzing long-term electronic health records (EHRs) of patients with ALS, the team can identify drugs or combinations of drugs prescribed for other conditions that may affect disease progression. The “untargeted” effect of drugs not only affects patient survival, but may also provide insight into how neurodegenerative diseases work and inform better treatments.
“When you talk to your ALS caregivers, you move because the illness has such a severe prognosis and you can do something,” said Priyadip Ray, staff scientist at LLNL's Department of Computational Engineering (CED).
Computer to the clinic
The Center for Disease Control estimates that as many as 31,000 Americans suffer from ALS (also known as Lou Gehrig's disease), and veterans are diagnosed at a higher percentage than the average population. This disease attacks motor neurons in the spinal cord and brain, increasing the loss of mobility until the body is closed, usually within 2-5 years. The cause is unknown and there is no treatment. Additionally, three FDA-approved drugs have minor effects.
However, the emergence of EHRS – a digital file with patient history, prescriptions, demographic information and more – has opened the door to unprecedented research opportunities.
Because ALS is a relatively rare disease and is rapidly developing, there is actually no number or time for large clinical trials to be conducted. [EHR] Data is important. This is because advanced AI/ML tools can be used to create excellent confidence hypotheses and perform 1-3-pair targeted clinical trials with much higher success rates. ”
Priyadip Ray, staff scientist at LLNL's Computational Engineering Division
In clinical trials, similar groups of patients will be given randomly either treatment or placebo. If half of the treatment has better results, it proves that the treatment works. Using EHR data, Ray and his team use a technique called causal machine learning.
“Causal learning creates some kind of synthetic clinical trial,” he said. “We looked for patients who were given a certain drug and matched a very similar group of patients and were likely but not likely to be given that drug.”
(Re) Move for purpose
His CED colleagues, Ray, Braden Soper, Andre Goncalves, Jose Cadena Pico and their collaborators, began by creating a surrogate model (mathematical approximation) of ALS progression using a small, publicly available EHR dataset. Through seed funding from the ALS Cure project, founded in memory of his wife by LLNL employee Mike Piscotty, the team has access to more than 20,000 EHR veterans with Veteran Affairs (VA) ALS. After the EHR scrubed the individual information, the team investigated the risk factors for ALS and received funding from the Department of Defense for further analysis.
The team studied 162 drugs patients regularly employ onset of ALS, and identified three classes that have a significant positive effect on survival: statins (reduce cholesterol), alpha blockers (reduce blood pressure and relieve muscles), and PDE5 inhibitors (treat erectile dysfunction). They also found that combining statins with alpha blockers was synergistic.
The team has discovered several early stage studies on these drugs and ALS, backing up the results, suggesting that they could all be good candidates. Stanford and UCLA collaborators conducted protein-protein interaction studies on each drug type and discovered several common downstream protein targets – the drugs ultimately had an impact.
“We are very excited about these initial discoveries,” Ray said. “If we can identify these shared downstream protein targets, we can target these proteins to create drugs that function even better.”
As VA data is heavily skewed towards men with military backgrounds (both risk factors for ALS), the team aims to support and generalize the results. To do this, they plan to analyze millions of patient files from the Optum EHR dataset. This was accessible thanks to new funding from the ALS Network, The ALS Cure Project, The Livermore Lab Foundation, RDM Positive Impact Foundation and Stanford University. They also plan to apply an AI/ML approach to study Parkinson's disease. Ray hopes to shed light on the treatment of all neurodegenerative diseases.
Meanwhile, the team is seeking funding to verify the findings of the clinical setting. This is not only one of the final steps to approving the drug for treatment of ALS, but also ensures that the approach works.
Ray is grateful for the opportunity to use AI/ML to make a difference in medical research and the unique infrastructure and connections with academia, industry and government that makes it possible.
“The lab recognizes that building these tools and using patient data can have a significant impact,” he said. “The opportunities to change in healthcare and national security have motivated me to tackle this influential research.”
sauce:
Lawrence Livermore National Laboratory (LLNL)
