Fair splits flip the leaderboard: CHANRG reveals limited generalization in RNA secondary-structure prediction
This work addresses the need for reliable RNA structure prediction in transcriptome annotation and RNA therapeutic design, but it is incremental as it focuses on improving evaluation rather than proposing a new method.
The study tackled the problem of overestimated generalization in RNA secondary-structure prediction by introducing CHANRG, a benchmark of 170,083 structurally non-redundant RNAs, which revealed that foundation-model methods lost most accuracy out of distribution while other methods remained more robust.
Accurate prediction of RNA secondary structure underpins transcriptome annotation, mechanistic analysis of non-coding RNAs, and RNA therapeutic design. Recent gains from deep learning and RNA foundation models are difficult to interpret because current benchmarks may overestimate generalization across RNA families. We present the Comprehensive Hierarchical Annotation of Non-coding RNA Groups (CHANRG), a benchmark of 170{,}083 structurally non-redundant RNAs curated from more than 10 million sequences in Rfam~15.0 using structure-aware deduplication, genome-aware split design and multiscale structural evaluation. Across 29 predictors, foundation-model methods achieved the highest held-out accuracy but lost most of that advantage out of distribution, whereas structured decoders and direct neural predictors remained markedly more robust. This gap persisted after controlling for sequence length and reflected both loss of structural coverage and incorrect higher-order wiring. Together, CHANRG and a padding-free, symmetry-aware evaluation stack provide a stricter and batch-invariant framework for developing RNA structure predictors with demonstrable out-of-distribution robustness.