Summary
A bioRxiv preprint describes DS3dRNA, a method that scores interactions among groups of RNA structural elements to design sequences for target structures. The authors report improved benchmark performance and experimental activity in designed Mango II and twister RNAs.
A bioRxiv preprint introduces DS3dRNA, a framework for designing RNA sequences around target three-dimensional structures. The method combines a statistical score for three-way interactions with physics-guided sequence sampling, and can account for multiple conformations. The authors report stronger results than representative RNA sequence-design methods on their evaluated benchmarks, alongside experimental activity in designed Mango II and twister RNAs.
RNA sequence determines how a molecule folds, and its shape helps determine what it can do. Predicting or designing a sequence from a desired structure is challenging because folding depends not only on pairs of complementary bases, but also on cooperative interactions involving more than two structural elements. DS3dRNA is designed to include these higher-order interactions in its scoring.
How DS3dRNA approaches RNA design
In RNA inverse design, the target is a structure and the task is to find a sequence that can adopt it. DS3dRNA combines a three-body statistical potential—a scoring function that evaluates interactions among three components—with physics-guided sampling of candidate sequences. The framework also supports design against multiple conformations, rather than requiring a single fixed structural target.
The researchers compared DS3dRNA with representative inverse-design methods. Across the benchmarks they evaluated, they report higher native-sequence recovery and closer agreement between predicted structures and the target structures. Native-sequence recovery measures how closely a designed sequence matches a reference sequence; structural agreement assesses whether the predicted fold resembles the intended one. The abstract also reports that lower design energies generally coincided with higher sequence recovery and macro-averaged F1 scores, a summary measure of classification performance.
Tests in functional RNAs
The team experimentally tested designs for Mango II, a fluorogenic RNA whose binding to a fluorescent dye produces a signal, and for the twister ribozyme, an RNA that catalyses its own cleavage. The authors report that the Mango II designs retained high-affinity fluorogenic activity. For five twister designs, mean endpoint cleavage fractions ranged from 37.7% to 50.6%, compared with 23.5% for the wild type.
These experiments provide functional tests alongside computational benchmark comparisons: the designed sequences were assessed not only for predicted structural agreement, but also for activity in two RNA systems. The results suggest that explicitly scoring higher-order interactions can complement other approaches to RNA design. The reported tests cover the RNA types described in the preprint abstract, so the results should be understood in that scope.
The work is a preprint posted on bioRxiv and has not yet undergone peer review.