Deep Learning-Guided Engineering of Bst DNA Polymerase Improves LAMP-Based Detection of Foodborne Pathogens

Loop-mediated isothermal amplification (LAMP) is a widely used nucleic acid detection method, but its application is often limited by the suboptimal performance of wild-type Bacillus stearothermophilus (Bst) DNA polymerase. This study employed a combined deep learning and semi-rational design strategy to engineer Bst DNA polymerase. High-throughput screening identified the A0A150MFP3 sequence and the L105M mutation, which increased enzymatic activity by 32.92%. Fusion with the CL7 protein generated a CL7-Bst mutant with enhanced thermal stability and tolerance to common inhibitors, including 7% (v/v) ethanol, 0.18‱ (w/v) SDS, 80 mmol/L NaCl, and 0.8 mmol/L EDTA. Systematic optimization of the LAMP reaction system determined the optimal pH (9.0), enzyme concentration (0.20 U/μL), and temperature (64 °C). When applied to Escherichia coli O157:H7 detection, the CL7-Bst mutant achieved Tt values of 15.13 and 12.78 for crude and purified DNA, respectively, with a limit of detection of 1 × 10³ CFU/mL. In summary, integrating deep learning with semi-rational design and fusion protein engineering yielded a high-performance DNA polymerase that facilitates rapid, sensitive, and field-deployable LAMP-based pathogen detection.