Chinese university releases open-source RNA model structRFM in Nature

Researchers from the University of Science and Technology of China have released an open-source RNA foundation model called structRFM. The findings were published in the journal Nature. Led by Professor Zhou Shaohua and Researcher Xiong Peng, the model improves RNA structure prediction and function analysis, and may help speed up drug discovery and other research.

RNA plays a key role in protein synthesis and gene regulation, and its function is closely tied to its sequence and folded structure. The team built about 21 million pairs of RNA sequences and structures to train the model. With just a sequence as input, it can pull out rich features and hits top-level results in many tasks.

Tests show structRFM works well in structure prediction and RNA classification tasks. The lab has opened up its full data, model weights, and source code. This is the university's 16th paper in top journals this year as a lead institution.