Fig. 2: In vitro demonstration and optimization of THETA.

a, A simplified scheme of THETA 3.9.9. Modifications from the original design (shown in Fig. 1b) are marked in teal. b, The increase of acetyl-CoA production from THETA version 1.0 to 3.9.9. Optimization was achieved through rational design (THETA 1.0 to 3.0) and machine learning-guided approaches (THETA 3.9.9). Manually confirmed test results are shown. c, Optimization of the THETA cycle in vitro by active learning with 30 different conditions in triplicates tested for each of the nine rounds. Assays (10 µl) were pipetted by an Echo liquid handler, started with 200 µM pyruvate and stopped after 3 h. Formed acetyl-CoA was quantified by LC–MS. As a control, V3.0 (purple circle, same composition as in the manual experiment shown in b) was prepared through the same procedure and is shown in comparison with the best THETA version (3.9.9, teal circle, same composition as in the manual experiment shown in b). d, Manually repeated experiments (120 µl scale) for V3.9.9 started with the original (200 µM pyruvate) and two more substrate concentrations. Samples were withdrawn at different timepoints and quantified by LC–MS. Data in b and d represent the mean ± s.d. obtained in n = 3 independent experiments except for THETA 1.0 and 2.0 in b (where data represent the mean of n = 2 independent experiments). Data in c represent the median with interquartile range derived from n = 30 different testing conditions, with each individual data point representing the mean obtained in n = 3 independent experiments. G6P, glucose-6-phosphate; 6PG, 6-phosphogluconolactone; G6PDH, glucose-6-phosphate dehydrogenase; CP, creatine phosphate; Cpk, creatine phosphokinase; Adk, adenylate kinase; DHO, dihydroorotate; Pyr, pyruvate.