Intelligent Design of <i>Escherichia coli</i> Terminators by Coupling Prediction and Generation Models

Terminators are specific nucleotide sequences located at the 3' end of a gene and contain transcription termination information. As a fundamental genetic regulatory element, terminators play a crucial role in the design of gene circuits. Accurately characterizing terminator strength is essential for improving the precision of gene circuit designs. Experimental characterization of terminator strength is time-consuming and labor-intensive; therefore, there is a need to develop computational tools capable of accurately predicting terminator strength. Current prediction methods do not fully consider sequence or thermodynamic information related to terminators, lacking robust models for accurate prediction. Meanwhile, deep generative models have demonstrated tremendous potential in the design of biological sequences and are expected to be applied to terminator sequence design. This study focuses on intelligent design of <i>Escherichia coli</i> terminators and primarily conducts the following research: (1) to construct an intrinsic terminator strength prediction model for <i>E. coli</i>, this study extracts sequence features and thermodynamic features from <i>E. coli</i> intrinsic terminators. Machine learning models based on the selected features achieved a prediction performance of <i>R</i>² = 0.72. (2) This study employs a generative adversarial network (GAN) to learn from intrinsic terminator sequence training data and generate terminator sequences. Evaluation reveals that the generated terminators exhibit similar data distributions to intrinsic terminators, demonstrating the reliability of GAN-generated terminator sequences. (3) This study uses the constructed terminator strength prediction model to screen for strong terminators from the generated set. Experimental verification shows that among the 18 selected terminators, 72% exhibit termination efficiencies greater than 90%, confirming the reliability of the intelligent design approach for <i>E. coli</i> terminators. In sum, this study constructs a terminator strength prediction model and a terminator generation model for <i>E. coli</i>, providing model support for terminator design in gene circuits. This enhances the modularity of biological component design and promotes the development of synthetic biology.