Machine Learning-Guided Engineering of T7 Rna Polymerase and Mrna Capping Enzymes for Enhanced Gene Expression in Eukaryotic Systems

Abstract

The integration of synthetic biology tools into eukaryotic systems offers both significant opportunities and challenges, particularly in optimizing transcriptional and post-transcriptional processes. T7 RNA polymerase (T7 RNAP) and mRNA capping enzymes (CEs) have been fused to enable eukaryotic mRNA production within a single construct. However, the activity of the fusion construct between the African Swine Fever Virus capping enzyme (ASFVCE) and T7 RNAP was relatively low. To address this, we fused the Brazilian Marseillevirus capping enzyme (BMCE) to T7 RNAP and developed a machine learning (ML) pipeline to engineer greatly improved fusion variants. This approach enabled the additive integration of nine predicted single substitutions that improved gene expression in yeast, thereby generating fusion polymerases that exhibited over 10-fold improvements in gene expression efficiency relative to the original fusion enzyme. Not only were ML substitutions additive for gene expression, they could be further combined with variants identified via directed evolution for even higher activities. By allowing ML predictions to guide validations we could rapidly explore the sequence landscape for enzyme optimization, achieving superior results even when compared to directed evolution. The improved enzymes have potential impact for numerous synthetic biology applications, including metabolic engineering, mRNA therapeutics, and cell free systems.

Keywords: T7 RNA Polymerase (T7 RNAP), mRNA Capping Enzyme (CE), Machine Learning (ML), protein engineering, Yeast Expression System, Synthetic Biology