Increasing mRNA Stability With CODEx

Overview

In protein-coding DNA, 61 sense codons encode 20 standard amino acids, allowing multiple synonymous codons for the same protein sequence. Codon selection influences translation rate, ribosome efficiency, and protein yield, which is critical in recombinant vaccine production where foreign genes must be efficiently expressed in host cells. This project presents a deep learning–based codon optimization system designed to improve translation efficiency. A convolutional neural network (CNN) was trained on approximately 3,900 yeast coding sequences, along with synthetic synonymous variants, to learn biologically meaningful codon usage patterns. The dataset was split into training, validation, and test sets, with approximately 15% (~580 sequences) reserved for independent testing. Model parameters, including weights, filters, and biases, were tuned to capture sequence-level patterns while avoiding overfitting. A constrained optimization algorithm performs synonymous substitutions while preserving the amino acid sequence. On the test set, optimized sequences achieved an average improvement of approximately 144% in the model’s optimization score compared to the original sequences. By improving codon usage patterns at the sequence level, this approach has the potential to increase recombinant protein yield, reduce production time, and lower manufacturing costs in vaccine and biotechnology applications.

Competition history

• CSEF 2026 Mathematical Sciences (Junior Division) · Entry J-14-12

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Mathematics projects
• CSEF Mathematics projects
• Every CSEF 2026 project