An interpretable RNA foundation model for exploring functional RNA motifs in plants

Area of Science:

• Plant Biology
• Genomics
• Bioinformatics

Background:

• Plant RNA plays a critical role in regulating growth, development, and stress adaptation.
• Foundation models (FMs) show promise in deciphering complex biological data.
• Existing models lack plant-specific RNA analysis capabilities.

Purpose of the Study:

• To introduce PlantRNA-FM, a high-performance, interpretable RNA foundation model tailored for plants.
• To enhance the understanding of plant RNA regulatory elements and their functions.
• To enable the identification of functional RNA sequence and structure motifs.

Main Methods:

• Pretraining PlantRNA-FM on a large dataset of RNA sequences and structures from 1,124 plant species.
• Utilizing an interpretable framework for motif identification.
• Conducting experimental validations to confirm findings.

Main Results:

• PlantRNA-FM achieved a superior F1 score of 0.974 for genic region annotation, significantly outperforming the previous best model (0.639).
• Identified biologically functional RNA sequence and structure motifs, including secondary and tertiary structures.
• Revealed translation-associated RNA motifs and their positional importance within genic regions.

Conclusions:

• PlantRNA-FM offers advanced capabilities for exploring functional RNA motifs in plant transcriptomes.
• The model empowers plant scientists to better understand and potentially program RNA codes.
• This work advances the field of plant RNA biology and bioinformatics.