Our AI-based phenotypic platform uniquely combines big data analysis and drug design. It can extract key performance markers from high-dimensional phenotypic readouts – such as shape, motility, behaviour and other observable characteristics – which are then used to generate and select new compounds for synthesis and assay.
Through this approach we can rapidly evolve novel compounds that satisfy key performance criteria. The same system can also be used to reveal whether compounds might interact with other well-researched protein targets or compounds that have been previously assayed by traditional target-based techniques or to specifically avoid well-trodden or undesirable mechanisms.
Application to phenotypic endpoints rather than traditional target-based work enables application to complex diseases where the target landscape is not sufficiently understood. The success of our Active Learning systems with phenotypic data was reviewed in Nature [need hyperlink].