Generative AI-driven artificial DNA design for enhancing inter-species gene activation and enzymatic degradation of PET

Abstract

Conventional approaches to heterologous gene expression rely on codon optimization, which is limited to swapping synonymous codons and often fails to capture deeper adaptive changes. In contrast, naturally evolved orthologous genes between species often differ by more than just synonymous substitutions - they can include non-synonymous mutations, insertions, and deletions that confer functional adaptation to different host contexts. Here we present OrthologTransformer, a Transformer-based deep learning model that converts orthologous genes between species by learning from large-scale orthologous gene datasets curated for high-quality sequence alignments. The model recapitulates the full spectrum of evolutionary differences - from synonymous codon swaps to amino acid-changing mutations and indels - to predict a coding sequence optimized for a target species while preserving the proteins function. In extensive validation across diverse bacterial species pairs, OrthologTransformer significantly increased the conversion accuracy of generated genes to native target sequences compared to the original source genes, even for pairs with stark disparities in GC content and optimal growth temperature. The Transformers context-aware designs also favored conservative amino acid substitutions, maintaining protein functional integrity. As a proof-of-concept, an OrthologTransformer-designed PETase gene for Bacillus subtilis from the host species Ideonella sakaiensis was synthesized and expressed, yielding robust PET plastic-degradation activity that surpassed synonymous codon-optimized controls. These results establish OrthologTransformer as a powerful tool for de novo cross-species gene adaptation, transcending the limits of traditional codon optimization and enabling more effective heterologous gene performance in synthetic biology applications.