PlantRNA-FM: An RNA Foundation Model for Exploration Functional RNA Motifs in Plants

The complex language of plant RNA encodes a vast array of biological regulatory elements that orchestrate crucial aspects of plant growth, development, and adaptation to environmental stresses. Recent advancements in large-scale pre-trained foundation models (FMs) have demonstrated their immense potential to decipher the complex 'language of nucleotide sequences'. In this study, we introduced PlantRNA-FM, a novel interpretable FM specifically designed for plant RNA research. PlantRNA-FM was pre-trained on an extensive dataset integrating RNA sequences and RNA structure information from 1,124 diverse plant species. It demonstrated superior performance in plant-specific downstream tasks, such as plant RNA annotation prediction and plant RNA translation efficiency (TE) prediction, compared to the second-best models, with F1 score improvements of 52.45% and 15.30% on the A. thaliana dataset, and 43.90% and 13.83% on the O. sativa dataset, respectively. In addition, we found that removing the RNA structural information from the model resulted in a performance decrease of up to 8.03%. Beyond a predictive model, PlantRNA-FM advances our understanding of plant RNA biology by identifying biologically functional RNA structural motifs, including canonical RNA structure motifs and RNA G-quadruplexes. We identified RNA structural motifs within the 5' UTR that are associated with high and low TE and conducted experimental validation. Disruption of the high-TE-associated RNA structures significantly reduced TE by 2.8-fold, whilst disruption of the low-TE structural motifs enhanced TE by 2.7-fold. Moreover, we identified RNA G-quadruplexes associated with low TE, and disruption of these structures led to a 3.3-fold increase in TE. By integrating large-scale pre-training with plant-specific data, PlantRNA-FM identify es functionally relevant RNA structural motifs, thereby deepening our understanding of the molecular mechanisms underlying plant growth, development, and stress responses. This study showcases the potential of PlantRNA-FM in unveiling the intricate regulatory networks encoded within plant RNA sequences and structures.