Comprehensive Datasets for RNA Design, Machine Learning and Beyond

This dataset addresses the need for a new community-wide standard benchmark specifically designed for RNA design and RNA modeling algorithms based on our previous experience. It is a very large, comprehensive, and general-purpose dataset containing over 15 million secondary structures with lengths ranging from 7 to 10,098. Our focus was mainly on multi-branched loops, which are often challenging to predict accurately. Consequently, this dataset encompasses a diverse range of complex motifs, from internal loops to n-way junctions (loops with n outgoing helices, where n ≥ 3), all extracted from RNA structures available in the RNAsolo and Rfam databases.