An Evolutionary Approach for Designing Stable and Highly Expressible Low-Immunogenicity Therapeutic mRNA Sequences

2026-05-27 · Source: Takara TLDR - Daily AI Papers · Field: Science & Research — Life Sciences & Biology, Mathematics & Computational Sciences, Artificial Intelligence & Machine Learning · Depth: Expert, medium

Summary

A novel two-stage in-silico framework integrates deep learning and evolutionary computation to optimize therapeutic mRNA sequences for stability, expressibility, and low immunogenicity. The first stage employs a pretrained CodonTransformer, a BERT-like Large Language Model, to generate biologically coherent mRNA sequences encoding a target antigen. Subsequently, a genetic algorithm refines these candidates through codon-aware crossover and synonymous mutation, guided by human codon usage preferences. Fitness functions evaluate translation metrics like CAI (achieving 0.73-0.74) and tAI (0.63-0.64), mRNA structural stability (global MFE of -346 to -356 kcal/mol, 84% base-paired), and reduced immunogenicity (average immune penalty lowered to 27.3). This BERT-GA framework improves CAI and tAI by over 6% and achieves a high codon-pair bias of 0.97, along with an unpaired_30 fraction of 0.87 for ribosomal accessibility. Unlike methods that prioritize extreme stability or translation efficiency alone, this approach balances these critical properties.

Key takeaway

For research scientists designing therapeutic mRNA sequences, you should consider integrating deep learning with evolutionary computation to achieve optimal balance across critical properties. Your current methods might over-optimize for single traits like extreme stability or translation efficiency, potentially compromising overall therapeutic efficacy. Adopt a multi-objective optimization approach, like the BERT-GA framework, to simultaneously enhance translation, structural stability, and reduce immunogenicity in your mRNA designs.

Key insights

The BERT-GA framework optimizes therapeutic mRNA sequences for balanced translation efficiency, structural stability, and low immunogenicity.

Principles

• mRNA design needs balanced stability and translation.
• Evolutionary algorithms refine deep learning outputs.
• Multi-objective fitness functions guide optimization.

Method

A two-stage in-silico framework: CodonTransformer generates initial sequences, then a genetic algorithm evolves them using codon-aware crossover and synonymous mutation, guided by multi-objective fitness functions.

In practice

• Combine LLMs with GAs for sequence optimization.
• Use CAI, tAI, MFE, and motif frequency as fitness metrics.
• Prioritize balanced properties over single-objective extremes.

Topics

• Therapeutic mRNA Design
• Codon Optimization
• Deep Learning
• Evolutionary Algorithms
• Immunogenicity Reduction
• mRNA Structural Stability

Code references

• ErikHartman/bopep

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• DualGPT-AB: a dual-stage generative optimization framework for therapeutic antibody design — DualGPT-AB optimizes therapeutic antibody design by combining conditional GPT with reinforcement learning for multi-property sequence generation.
• Convolutional neural networks quantify antibiotic resistance in Mycobacterium tuberculosis with diagnostic grade accuracy and predict treatment response — CNNs can predict antibiotic MICs from genomic data with clinical accuracy, leveraging diverse biological features.
• Generalizable mutation-effect prediction across adaptive immune recognition via unified multimodal framework — UniAIR unifies multimodal data and structural information to predict immune mutation effects with high generalizability.
• AgForce Enables Antigen-conditioned Generative Antibody Design — Existing antibody design models fail to condition on antigen, suffering from antigen blindness and vocabulary collapse due to cross-entropy loss.
• Bringing AI-driven protein-design tools to biologists everywhere — Democratizing AI-driven protein design tools empowers biologists to accelerate drug discovery and engineer novel proteins.
• AWS and Hopkins Engineering announce groundbreaking database for AI/ML antibody design — A new diverse, experimentally validated dataset enables rigorous benchmarking of AI models for antibody developability.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.