Optimizing Biomanufacturing 1,2,4-butanetriol with JMP Hybrid DOE and SVEM (2026-US-PO-2976)

Biomanufacturing 1,2,4-butanetriol (BT) has been of interest due to its utility as a precursor. Efforts to engineer cells to produce BT have had mixed success. In this project, we explored the production of BT with cell-free metabolic engineering, where cell lysates containing each of the enzymes needed to produce BT are mixed along with the precursor xylose as well as multiple cofactors. The design space for blending these components was very large.

Therefore, Mixture DOE was used to efficiently explore the design space. Due to the nonlinear interactions that are characteristic of biomanufacturing systems, it was determined that an I-optimal mixture DOE design using a typical Scheffé model might not be sufficient to model the system. Therefore, JMP Pro was used to build a hybrid DOE of I-optimal and space filling. Initially, a standard Scheffé prediction model was fit. Then, to take advantage of the space-filling hybrid design, JMP Pro was used to fit a SVEM (self-validating ensemble model) to capture any subtleties that might be eluding the Scheffé model.

The JMP Multi-Objective Optimization Profiler was used to predict formulations that would simultaneously maximize BT production while minimizing cost. Surprisingly, some of the predicted optimal formulations did not require the most expensive formulation components that were previously believed to be essential. As confirmation runs would later confirm, the DOE predicted optimal formulation doubled the BT production, compared to the accepted reference formulation, and, at the same time, reduced the cost by 100X.