Fig. 1: Machine learning to guide the design of high-activity combinatorial variants. | Nature Communications

Fig. 1: Machine learning to guide the design of high-activity combinatorial variants.

From: Machine learning-guided evolution of pyrrolysyl-tRNA synthetase for improved incorporation efficiency of diverse noncanonical amino acids

Fig. 1

a Structure of IFRS used for engineering TBD of PylRS, and 3BrF was selected as the substrate of IFRS. b Dataset 1 is composed of activities of 12 single-point mutants. c Dataset 2 is composed of activities of double and triple mutants used as a test set. d Accuracy of ML model in Dataset 2. e Dataset 3 is composed of activities of quadruple and multiple-point mutants used as a test set. f Accuracy of ML model in Dataset 3. g The SCS efficiency of the top 8 variants predicted by the ML model. h Experimental data and predicted data presented in a 2-dimensional sequence space. Error bars represent ±standard deviation of the mean over three independent replicates. Source data are provided as a Source Data file.

Back to article page