New machine learning algorithm for drug discovery that is twice as efficient as the industry standard and identified potential ligands for the M1 receptor, a potential target for the treatment of Alzheimer’s disease.
A paper from Lee et al.  (University of Cambridge) in collaboration with Pfizer, describes the development of an algorithm to use machine learning to separate pharmacologically relevant chemical patterns from irrelevant ones. The algorithm compared active versus inactive molecules at the muscarinic acetylcholine receptor, M1 and uses machine learning to identify components of the compounds are important for binding and which arose by chance. Lee et al. built a model using historic data using ~5,000 compounds that were screened for agonist activity, of which 222 were active. Six million molecules from the e−Molecules database were computationally screened. From these ~100 molecules were purchased and screened in CHO cells expressing the M1 receptor with four compounds identified as agonists (EC50 range 80-300nM).
Comments by Anthony Davenport, IUPHAR/BPS Guide to PHARMACOLOGY, University of Cambridge
(1) Lee AA et al. (2019). Ligand biological activity predicted by cleaning positive and negative chemical correlations. PNAS, https://doi.org/10.1073/pnas.1810847116. [Epub ahead of print]. [PNAS: Article]