Using a series of machine learning models and laboratory experiments, the researchers discovered a set of FDA-approved drugs that could provide lipid-lowering properties to patients dealing with high cholesterol, according to the research that was screened. Acta Pharmacolica Sinica. If reuse is successful, these drugs could expand the arsenal of lipid-lowering therapy available to clinicians and pharmacists to provide to patients.1,2
New lipid-lowering agents can support those intolerant of traditional statin therapy. | Image credit: ©Khukrit -Stock.adobe.com
The Important Need for New Drugs to Treat Hyperlipidemia
Hyperlipidemia has a high presence in the US population. The American Heart Association's 2023 Heart Disease and Stroke Statistics Update found a 34.7% prevalence of hypercholesterolemia, with about a quarter of US adults having elevated low-density lipoprotein cholesterol (LDL-C) levels between 2017 and 2022. For patients.1,3
A variety of lipid-lowering therapies are available to patients in the form of statins, cholesterol absorption inhibitors, and protentan converters subtilisin/kexin type 9 (PCSK9) inhibitors. These drugs, with their own mechanism of action, are effective in lowering lipid levels, but patients face many challenges with uptake of such drugs. Studies have shown that some individuals have less resistance to existing treatments, while others have less susceptibility to such medications.1,4,5
It is important to identify new strategies for lipid reduction. Machine learning is a promising path for such exploration, characterized by its ability to extract and identify critical features from biomedical datasets while determining possible drug disease associations. To determine new measures of lipid reduction, investigators sought to use machine learning to identify FDA-approved non-lipid lowering drugs that could be reused to help patients lower cholesterol.1
The authors aimed to apply machine learning methods using public databases and literature to summarize a set of drug candidates that could be reused to reduce lipid levels in patients. We have developed a model that can verify the lipid-lowering effects of non-lipid-lowering drugs using multiple machine learning algorithms. Additionally, to assess the validity of machine learning results, the authors conducted retrospective clinical data validation and animal experiments.1
Multiple drugs indicate potential for significant lipid reduction
A new machine learning framework was utilized to analyze 3430 drugs. 176 known lipid-lowering agents and 3254 controls. Following the analysis of the results, at least eight models identified a total of 29 FDA-approved drugs with no indication for lipid lowering that could have potential lipid lowering.1,2
The authors then sought to validate these drugs through retrospective clinical data analysis. Comparative analysis of the mean blood lipid profiles of patients before and after drug therapy, four drugs – agatrovan (ACOVA), levothyroxine sodium (leboxyl; Pfizer), oseltamivir (Tamiflu; Roche), and thiamine – a modelling of patients meaningful biological activities. Argatroban has been found to elicit the most prominent effects, such as LDL and triglycerides.
In vivo experiments conducted via mouse models showed other drugs that significantly modulate important lipid-related blood indicators. Again, sodium levothyroxine, along with sulfaphenazole, induced an important triglyceride-lowering effect. Six drugs, including Sorafenib (Nexavar; Bayer), Prasterone (Intrarosa; Cosette Pharmaceuticals), and Regorafenib (Stivarga; Bayer), showed significant effects on blood-density lipoprotein (HDL) levels. Prasterone showed the most prominent HDL elevator effect in the experiment.
A vast possibilities
The results of this trial not only provide numerous non-lipid lowering therapies that can actually be used to lower lipids, but also demonstrate the feasibility of machine learning in examining datasets, exploring drug possibilities, and incorporating numerous sources to provide solutions that can assist clinicians and pharmacists. The numerous validations received by investigators provide strong reliability in their findings, but interpreters still need to be careful given the limitations of artificial intelligence.1
All identified candidates have been found to be particularly pronounced across argatroban, sodium levothyroxine, and sulfaphenazole, and may lower lipids. As the authors argue, these agents may serve as useful gap fillers for patients who cannot tolerate or respond well to traditional lipid-lowering therapies. According to the authors, newly identified agents can be combined with existing agents to achieve synergistic lipid-lowering effects. Beyond drug reuse, the results may offer new directions for researchers in the development of targeted therapies to control lipid levels.1
“We have established a paradigm for AI-driven drug relocation,” study senior author Peng Luo, MD, said in a news release with the results. “Bypassing decades of traditional drug development by integrating computational predictions with clinical and experimental validation. It provides clinicians with new tools faster and cheaper.”2
