USTC Develops Structure-Guided RNA Foundation Model Featured in Nature
A team from the University of Science and Technology of China has introduced structRFM, a structure-guided foundation model for RNA. The paper was published online in *Nature* on November 5. The model incorporates base-pairing information from secondary structures into sequence pretraining, was trained on approximately 21 million sequence-structure pairs, and achieves state-of-the-art performance across multiple tasks. The team also developed the MUSES method to integrate various secondary…
USTC Develops Structure-Guided RNA Foundation Model Featured in Nature
A team from the University of Science and Technology of China (USTC) has proposed structRFM, a structure-guided RNA foundation model. The paper was published online in Nature on November 5. The model incorporates base-pairing information from secondary structures into sequence pretraining and was trained on approximately 21 million sequence-structure pairs, achieving state-of-the-art performance across multiple tasks.
The team also developed the MUSES method to integrate multiple secondary structure prediction results and reduce bias. Its 3D structure prediction module, Zfold, performs comparably to AlphaFold 3. The data, model weights, and code have all been made publicly available.
