Researchers at the University of Edinburgh use machine learning algorithm to identify drugs with potential new applications in minutes
ANALYSIS
Artificial intelligence (AI) has been the driving force behind a lot of big developments in the last year. But while super-intelligent chatbots and rapid art generation have gripped the internet, elsewhere AI has been used to try and find solutions to one of humanity’s biggest problems: ageing.
Researchers at the University of Edinburgh, working in the field of drug discovery, have used machine-learning systems to unearth a selection of new anti-ageing drugs.
Machine learning is a branch of AI that focuses on using data to imitate the way that humans learn, improving its accuracy as its fed more data. In the past, machine learning has been used to create chess-playing robots, self-driving cars and even Netflix recommendations, but in this case the algorithm was looking for senolytics.
Senolytics are drugs that are able to slow ageing, as well as prevent age-related diseases. They work by killing off senescent cells, which, although still alive, are no longer able to replicate. While having cells that don’t replicate isn’t necessarily a bad thing, they will have suffered damage to their DNA (sunburned skin cells, for example), so stopping replication stops the damage from spreading.
Vanessa Smer-Barreto, a research fellow at the University of Edinburgh’s Institute of Genetics and Molecular Medicine, was investigating new drugs, specifically senolytics, in her post-doctorate research.
Frustrated by the expense and time involved in the process of drug discovery, she turned to machine learning in the hopes of reducing both.
“Generating your own biological data can be really expensive, and it can take a lot of time, even just to gather training data,” she says.
“What made our approach different to others is that we tried to do it on limited funds. We took training data from existing literature and looked into how to use this with machine learning to speed things up.”
By using a machine learning algorithm, she was able to find three promising senolytics.
UNMISTAKABLE EXAMPLES
Smer-Barreto and her colleagues fed an AI model with examples of known senolytics and non-senolytics, teaching it to distinguish between the two. The AI could then be used to predict whether molecules it hadn’t seen before could be senolytics based on whether or not they matched the exam
