[P001] Poster

12:10 PM - 1:00 PM JST(3:10 AM - 4:00 AM UTC)

[P01-091] Accelerated DBTL Uncovers the Global Optimum from 2,187 Variants

○Yuyoung An^1,2, Wonjae Seong¹, Seong-Kun Bak^1,2, Haneul Kim¹, Hyoseok Ha¹, Dae-Hee Lee^1,2,3, Seung-Goo Lee^1,2,3, Haseong Kim^1,2,3 (1. KRIBB (Korea), 2. KAIST (Korea), 3. UST (Korea))

Keywords:

Synthetic Biology,DBTL,Biofoundry,Modular Assembly,Combinatorial Optimization,Metabolic Engineering,Mevalonate Pathway

Biology has evolved from a science into an engineering discipline aimed at constructing biological systems to meet societal demands. However, due to the nonlinear and complex nature of biological systems, predicting phenotypes from given inputs remains challenging and experimental design continues to be a major bottleneck in synthetic biology. Exploring all combinations of variables enables comprehensive landscape mapping and facilitates more efficient identification of global optima. The Design–Build–Test–Learn (DBTL) cycle is essential to accelerate this approach, but the Build and Test stages often limit overall throughput. Here, we developed an automated Build framework in a biofoundry based on hierarchical modular assembly. By eliminating time-consuming steps such as transformation and miniaturizing assembly reaction volumes, we achieved the economic efficiency required for scalable experimentation. For the high-throughput Test stage, we implemented colony-level quantitative color analysis, supported by colony detection using a CycleGAN-based deep learning model. As a proof of concept, we screened 2,187 expression combinations by independently regulating seven genes of the mevalonate (MVA) pathway as separate transcriptional units using three levels of ribosome binding site (RBS) strength, and identified patterns in metabolic flux distribution. This study demonstrates a practical implementation of a high-throughput DBTL cycle and presents a strategy to overcome longstanding bottlenecks in biological engineering.