Improving human protein production using advanced AI methods for genetic code optimization

COformer is a deep learning framework for codon optimization that enhances protein expression and mRNA stability.

• The model integrates convolutional neural networks and transformers to analyze codon-level features linked to protein expression in human cells.
• Traditional codon optimization methods are limited by their reliance on frequency-based heuristics and do not consider sequence context.
• COformer learns both local and global sequence contexts that may influence translation efficiency and stability.
• It is trained on sequences optimized by commercial tools and generates codon choices validated in in vitro experiments.

Simplified