Breaking Down Senolytics: AI and the Race to Ageing’s Finish Line
You might call them ‘zombie cells’, but don’t worry, no apocalyptic drama here. These are senescent cells – metabolically active, yet unable to replicate due to damaged DNA. Sounds like a Halloween spin-off, but this is the daily reality of our cells facing an onslaught of assaults, from UV rays to chemical exposure.
Now, don’t get me wrong, senescent cells aren’t the villains here. In fact, their inability to replicate could actually stop the damage from spreading. The real problem arises when they start throwing inflammatory protein parties, inviting the neighboring cells and turning our bodies into inflammatory wastelands. The link? Senescent cells have been implicated in a host of diseases like type 2 diabetes, COVID, pulmonary fibrosis, osteoarthritis, and, yes, the big C: cancer.
Enter stage left: senolytics – drugs that show these zombie cells the exit door. As promising as they sound, we have only managed to test a couple in humans. That’s where my colleagues and I come in, armed with the potent power of machine learning.
Turning to Tech: Machine Learning in the Drug Discovery Process
Let me walk you through the process. Imagine a room filled with researchers from the University of Edinburgh and the Spanish National Research Council IBBTEC-CSIC in Santander, Spain. Our mission: to see if we could teach AI to identify new senolytic drug candidates.
You might ask, how do you train AI? By feeding it examples of known senolytics and non-senolytics. We test various models, determining the best by who makes the fewest errors. We handed over 4,340 molecules to our chosen model and just five minutes later – voila – a list of 21 molecules likely to be senolytics.
Impact and Potential: Big Pharma, Meet AI
Now, I want you to imagine manually testing all 4,340 molecules in a lab. Think weeks of gruelling work, and a bill of around £50,000 just to buy the compounds. Enter our AI model, saving time, money, and probably a few headaches.
Next, we pitted these 21 candidates against two types of cells: healthy and senescent. Of the 21, three rose to the occasion, eliminating senescent cells without harming the healthy ones. Out of these three, one proved to be even more effective than the best-performing known senolytic.
This is what happens when data scientists, chemists, and biologists team up – a potential game-changer in how we find treatments and cures for diseases. Our trio of potential senolytics is now being tested in human lung tissue, so keep an eye out for updates.
