BERT-BASED GENERATION AND GENETIC ALGORITHM OPTIMIZATION OF SYNTHETIC ANTIGEN-ENCODING MRNA SEQUENCES

Messenger RNA (mRNA) therapeutics require optimized sequence design to ensure efficient translation, structural stability, and minimal immunogenicity. This study presents a two-stage in-silico framework that integrates deep learning and evolutionary computation for rational mRNA optimization.

In the first stage, a pretrained CodonTransformer (BERT-like Large Language Model) generates biologically coherent mRNA sequences encoding the target antigen. In the second stage, a genetic algorithm (GA) evolves these candidate sequences through codon-aware crossover and synonymous mutation guided by human codon usage preferences. Fitness functions for evaluation combined translation-related metrics (CAI, tAI, codon-pair bias), RNA structural stability (local and global MFE via RNAfold, GC content), and immunogenicity (CpG/UpA motif frequency).

Over successive generations (38th, 40th, and 42nd), the GA improved CAI and tAI by over 6% and codon-pair bias is high and consistent (~ 0.97), stabilized global MFE near −340 kcal/mol, and reduced immune-stimulatory motifs. The optimized constructs demonstrated balanced translation efficiency and structural accessibility, highlighting the proposed BERT–GA framework as an effective, data-driven approach for in-silico mRNA sequence design and optimizations

This therapeutic mRNA optimization framework or pipeline is particularly relevant for cancer mRNA therapeutics, where precise control of translation efficiency and structural stability is critical for robust antigen expression in tumor-targeted vaccines. The framework enables rapid generation of optimized mRNA constructs for neoantigen-based immunotherapy, supporting personalized cancer vaccine development.