Recommended

New Webinar: Neurological disorders – organoids for early drug discovery
Listen to our latest podcast: breaking resistance – the future of immunotherapy in oncology
Download now: The latest advancements and breakthroughs in immuno-oncology
Register now: Ion Channel Drug Discovery – Recent Advances in Novel Non-Opioid Pain Research
Register now: SAFIRE: AI predicted ADME for drug discovery: leveraging bioprint
article

Transformational machine learning approach could accelerate drug design

In this Q&A, Professor Ross King from the University of Cambridge, UK, discusses how a new machine learning approach could aid drug discovery and development. The method learns from multiple problems and improves performance while it learns.

Why are drug discovery researchers using machine learning to enhance therapeutic development?

Machine learning is the best way to predict drug activities; it has been advantageous to researchers for many years. For certain tasks, especially abstract tasks, machines are better than human beings and the empirical evidence is that machines are better at optimising several things at the same time.

You have developed a new transformational machine learning (TML) method. Can you give me a bit of background into how you developed this approach?

Traditionally in machine learning, you would represent the objects you are interested in learning about to the computer in a tabular form, with the columns representing the descriptors and the rows containing the examples. In drug design, people most commonly use molecular fingerprints to describe the structure of molecules, with each column representing a specific substructure. The machine learning then works by…