Illuminating enzyme design using deep learning

From humans designing machines, to machines designing biology, deep learning is turning the tables for assisted exploration of biologically active and diverse protein designs. Now, a deep-learning-based strategy has been used to design artificial enzymes that catalyse a light-emitting reaction.

Nature uses proteins called enzymes to catalyse the reactions necessary for life with greater selectivity and efficiency than most synthetic catalysts. However, natural enzymes do not catalyse many reactions of interest because they are not useful to living organisms, the reagents are not available, or the conditions required for the reactions do not exist in nature. While directed evolution can unlock new enzyme chemistries by making changes to existing enzymes, it requires a starting point with some activity for the desired reaction. De novo rational design overcomes this limitation by introducing active sites and substrate-binding pockets that are predicted to catalyse a reaction of interest into geometrically compatible scaffolds, but is limited by a lack of suitable protein structures and the complexity of native protein sequence–structure relationships. De novo designed enzymes often lack the activity and selectivity of natural enzymes.

留言 (0)

沒有登入
gif