RNAGEN: A Generative Adversarial Network-Based Model to Generate Synthetic RNA Sequences to Target Proteins.

Area of Science:

• Biochemistry
• Computational Biology
• Molecular Biology

Background:

• RNA-protein interactions are crucial for regulating protein function, localization, and stability.
• Targeting proteins with small molecules or other proteins is common, but designing small RNAs for this purpose remains underexplored.
• A key challenge is the lack of methods to generate RNA molecules with specific protein-binding capabilities.

Purpose of the Study:

• To develop a computational method for generating RNA sequences designed to bind specific target proteins.
• To leverage generative adversarial networks (GANs) and optimization techniques for RNA sequence design.
• To address the challenge of designing RNA molecules for targeted protein binding.

Main Methods:

• Utilized generative adversarial networks (GANs) to generate short RNA sequences with inherent RNA-like properties (e.g., GC content, free energy).
• Employed an optimization technique to fine-tune generated RNA sequences for specific protein binding.
• Integrated existing RNA-protein binding prediction models to guide the generative process, including using models trained on related proteins when a specific model was unavailable.

Main Results:

• Successfully generated piRNA sequences predicted to bind the SOX2 protein, using models trained on related proteins (SOX15, SOX14, SOX7).
• Experimentally validated in vitro that the top generated piRNA molecules specifically bind to SOX2.
• Demonstrated the efficacy of the generative model and optimization approach in producing high-affinity RNA binders.

Conclusions:

• The proposed generative model combined with gradient-based optimization effectively generates piRNA sequences with high predicted binding affinity for target proteins.
• The generated RNA sequences were validated using state-of-the-art RNA-protein binding prediction models.
• This method offers a promising approach for designing functional RNA molecules for targeted protein interactions.