GenerRNA: A generative pre-trained language model for de novo RNA design

Area of Science:

• Biotechnology
• Computational Biology
• Molecular Biology

Background:

• RNA design is vital for vaccines, therapeutics, and biotechnological tools.
• Current RNA design methods lack versatility and rely on predefined structures or knowledge.
• A need exists for adaptable and knowledge-independent RNA generation techniques.

Purpose of the Study:

• Introduce GenerRNA, a Transformer-based model for RNA sequence generation.
• Enable the creation of novel RNA sequences with stable secondary structures.
• Demonstrate fine-tuning GenerRNA for specific RNA functionalities, such as protein binding.

Main Methods:

• Leveraged large language models (LLMs) and Transformer architecture.
• Pre-trained GenerRNA on extensive RNA sequence datasets.
• Fine-tuned the model on specialized datasets for targeted RNA generation tasks.

Main Results:

• GenerRNA successfully generates novel RNA sequences with stable secondary structures.
• Generated sequences are distinct from existing ones, broadening RNA sequence space.
• Fine-tuned GenerRNA achieved the generation of RNA sequences with high target protein affinity.

Conclusions:

• GenerRNA represents the first generative language model application for RNA design.
• This innovative approach overcomes limitations of existing RNA design techniques.
• GenerRNA offers a versatile platform for advancing RNA-based biotechnologies.