Redesigning the Eterna100 RNA design benchmark for the Vienna 2 folding engine

RNA secondary structure design performance on Eterna100 changes when evaluated with ViennaRNA 2 instead of ViennaRNA 1, and both Eterna100-V1 and Eterna100-V2 remain unsolved under intended settings.

• Why it matters: RNA design depends on the inverse of RNA structure prediction, so benchmark difficulty can shift with the thermodynamic “folding engine,” risking misleading algorithm comparisons.
• What they did: An online community modified Eterna100 to create Eterna100-V2 for ViennaRNA 2, then six RNA design algorithms were trained and tested under the updated engine settings.
• The result: Prior “solved” results for original Eterna100 were linked to incorrect folding engine settings, implying that future algorithms may need to adapt to updated thermodynamic models.

AI summary of the abstract · may contain errors · not medical advice · read the paper.