CodonMamba: a foundation model for programmable mRNA coding sequence design
Although mRNA codon language models provide a generalizable framework for biological sequence design, effective CDS design requires both a learned sequence design space that captures biological constraints and context-configurable design preferences. Here we present CodonMamba, a codon language model framework for mRNA prediction and programmable CDS design. Pretrained on large-scale…
CodonMamba is a groundbreaking codon language model framework designed for mRNA prediction and programmable coding sequence (CDS) design. Trained on extensive coding-sequence data, CodonMamba outperformed existing models in 10 out of 12 benchmark mRNA prediction tasks. What sets CodonMamba apart is its ability to transform codon optimization from a trained-inference-dependent process to an inference-time steerable generation system.
By allowing users to input codon usage preferences during generation, CodonMamba facilitates context-specific mRNA sequence engineering without the need for retraining. In experimental evaluations, CodonMamba demonstrated the capability to coordinate multiple design-relevant sequence properties and enable programmable cross-host CDS retargeting through simple prior switches during inference.
This innovative approach preserves the majority of model-derived codon selections while offering unprecedented flexibility for precise mRNA design using foundation models.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
