The process of finding new drugs, known as drug discovery, has traditionally been a costly and time-consuming endeavor. However, machine learning, a form of artificial intelligence, has emerged as a game-changer that can significantly accelerate this process at a fraction of the cost.
In a recent breakthrough, a team of researchers including colleagues from the University of Edinburgh and the Spanish National Research Council IBBTEC-CSIC in Santander, Spain, harnessed the power of machine learning to identify three highly promising candidates for senolytic drugs. Senolytics are a class of drugs that slow down the aging process and help prevent age-related diseases.
Understanding Senolytics and Their Role:
Senolytics function by eliminating senescent cells, commonly referred to as "zombie cells." These cells remain metabolically active but lose their ability to replicate due to DNA damage caused by various factors, such as exposure to UV rays. While the inability to replicate can halt the spread of damage, senescent cells can also release inflammatory proteins that affect neighboring cells. Accumulation of senescent cells has been associated with a range of diseases, including type 2 diabetes, COVID-19, pulmonary fibrosis, osteoarthritis, and cancer.
Laboratory studies on mice have demonstrated that senolytics can mitigate these diseases by selectively targeting and eliminating senescent cells while preserving healthy cells.
Machine Learning Accelerates the Discovery Process:
With approximately 80 known senolytic drugs, only two have been tested in humans thus far, namely dasatinib and quercetin. The quest for additional senolytics that can address diverse diseases typically entails a lengthy timeline of 10 to 20 years and requires substantial financial investment.
To expedite this process, the research team set out to train machine learning models capable of identifying new senolytic drug candidates. They fed the models with data on known senolytics and non-senolytics, enabling the models to learn the distinguishing characteristics between the two. Subsequently, these models could predict whether previously unseen molecules possessed senolytic properties.
The Best Model and Swift Results:
To determine the most effective model, the team evaluated different machine learning models by testing them on a portion of the training data that was kept hidden until the training process concluded. This approach allowed them to quantify the models' error rates and select the best-performing one.
Once the optimal model was established, it was deployed to make predictions. The team fed the model 4,340 molecules, and within a mere five minutes, it provided a list of results.
The AI model identified 21 top-scoring molecules with a high likelihood of being senolytics. In a conventional laboratory setting, testing these 4,340 molecules would have required weeks of intensive work and incurred costs of £50,000, not accounting for experimental equipment and setup expenses.
Validation and Promising Candidates:
The team proceeded to test the identified drug candidates on healthy and senescent cells. Out of the 21 compounds, three—periplocin, oleandrin, and ginkgetin—successfully eliminated senescent cells while preserving the majority of normal cells. Further investigations into these newly discovered senolytics were conducted to gain deeper insights into their mechanisms of action.
Detailed biological experiments revealed that among the three drugs, oleandrin exhibited greater effectiveness than the best-performing known senolytic drug of its kind.
The Potential Impact of an Interdisciplinary Approach:
The interdisciplinary collaboration between data scientists, chemists, and biologists has paved the way for significant advancements. With sufficient high-quality data, AI models have the potential to accelerate the remarkable work carried out by chemists and biologists, especially in identifying treatments and cures for diseases with unmet needs.
Future Research and Human Trials:
Having validated the efficacy of these senolytic candidates in senescent cells, the researchers have progressed to testing them on human lung tissue. The team anticipates sharing their next set of results within two years, marking another milestone in the journey toward combating aging and age-related diseases.
The utilization of machine learning in drug discovery represents a groundbreaking approach that promises to revolutionize the field. By harnessing the power of AI to identify potential senolytic candidates, researchers are able to expedite the process, reducing costs and opening up new possibilities for the treatment and prevention of aging and age-related diseases.
Comments
Post a Comment