RNA structure prediction improves when SHAPE data is modeled with Markov models.

RNA molecules fold into specific shapes to control how cells work. Scientists use experiments like SHAPE to measure which parts of an RNA chain are flexible and which parts pair up into stems. Computer programs use these experimental signals to predict an RNA molecule's shape. However, existing programs assume each measurement along the RNA strand is completely independent, ignoring real-world patterns between neighboring building blocks. To solve this, researchers converted SHAPE measurements into distinct steps and built Markov models, which are statistical tools that track patterns where one event depends directly on the one before it. These models recognize signature patterns in loops and stems, and they classify small structural shapes. When the researchers combined these models and specific signals from stable loop motifs into folding algorithms, the software predicted RNA shapes more accurately than current leading methods. This shows that accounting for nearby signals helps algorithms map RNA much better.

Why it matters

Researchers studying RNA's role in gene regulation and disease rely on structural predictions from chemical probing experiments. Current methods treat each measurement in isolation, missing important patterns between nearby positions. This work captures those patterns, leading to more accurate structure predictions, especially in loop regions where traditional methods struggle. Better RNA structures mean deeper insights into cellular processes and improved targets for drug design.

Questions answered