Latent-Space Codon Optimization for Enhanced Protein Expression
October 5, 2026
LSCO enables gradient-based codon optimization by mapping discrete sequences into the continuous latent space of a pretrained mRNA language model. The method integrates an uncertainty-aware expression predictor with a Minimum-Free-Energy regularizer and a protein-to-codon back-translation model to maximize translation efficiency.
HOW THIS AFFECTS YOU
●
researcherYou can replace heuristic proxies like GC-content with continuous latent space searches for more accurate protein design.
●
healthThis could accelerate the design of more effective mRNA vaccines and therapeutic proteins.