Fig. 4. One-step method to predict protein yield in cell-free systems.

a Illustration of the method used to predict the yield of protein expression with a new lysate, labeled PS, made by another experimentalist. The training of the model is based on yield measurements of 20 buffer compositions (magenta circles). The choice of the 20 buffers combinations leading to the best predictions is described in the methods. The yields obtained with 82 compositions (green circles) were measured and compared to the model predictions to test its accuracy (R² value was computed on 102 values, Q² on the 82 values of the test set, the linear regression fits the 82 test values). The yield is specific to each lysate as the reference composition used the same chemicals concentration as in Fig. 1 but with different lysate. b Comparison of the yields obtained with the lysate AB (made by a third experimentalist) vs. the model predictions. c Comparison of the yields obtained with the lysate, of panel a, supplemented with 0.25 mg/mL of novobiocin vs. the model predictions. d Comparison of the yields obtained with the lysate of a, supplemented with 0.5 mg/mL of spectinomycin vs. the model predictions. e Comparison of the yields obtained with a lysate obtained from the strain DH5a vs. the model predictions. In all panels, the model predictions are based on a model trained with the same 20 buffer compositions and the same test set of 82 buffer compositions only lysate differs. In all panels, the data are mean and the horizontal gray lines stand for the standard deviation of three replicates. The horizontal gray lines stand for the standard deviation of 25 predictions and the vertical gray lines correspond to the standard deviation obtained on three replicates. Source data are provided in the Source Data file.