The role of low-complexity repeats in RNA–RNA interactions and a deep learning framework for duplex prediction

RNA-RNA interactions (RRIs) are fundamental to gene regulation and RNA processing, yet their molecular determinants remain unclear. In this work, we analyzed several large-scale RRI datasets and identified low-complexity repeats (LCRs), including simple tandem repeats, as key drivers of RRIs. Our findings reveal that LCRs enable thermodynamically stable interactions with multiple partners, positioning them as key hubs in RNA-RNA interaction networks. RNA-sequencing of the interactors of the Lhx1os lncRNA allowed to validate the importance of LCRs in shaping interactions potentially involved in neuronal development. Recognizing the pivotal role of sequence determinants, we developed RIME, a deep learning model that predicts RRIs by leveraging embeddings from a nucleic acid language model. RIME outperforms traditional thermodynamics-based tools, successfully captures the role of LCRs and prioritizes high-confidence interactions, including those established by lncRNAs. RIME is freely available at https://tools.tartaglialab.com/rna_rna. Overall design: RNA sequencing of two independent Lh1xos pull-down experiments, which allowed the isolation of the Lhx1os lncRNA using 20 nt-long antisense biotinylated oligonucleotide probes in mESC-derived motor neurons. The experiments were performed employing two sets of probes, ODD and EVEN, targeting all the Lhx1os isoforms. A third set of probes, targeting the LacZ mRNA, was used as a negative control (LacZ). The enrichment in each pull-down sample was assessed by comparing it with the Input sample.