"RNA Profiling: Extracting Structural Signals from Noisy Distributions"
Christine Heitsch, Ph.D.
Professor, School of Mathematics
Accurate RNA structural prediction remains challenging, despite its increasing biomedical importance. Sampling secondary structures from the Gibbs distribution yields a strong signal of high probability base pairs. However, identifying higher order substructures requires further analysis. Profiling (Rogers & Heitsch, NAR, 2014) is a novel method which identifies the most probable combinations of base pairs across the Boltzmann ensemble. This combinatorial approach is straightforward, stable, and clearly separates structural signal from thermodynamic noise. Moreover, it can be extended to predict consensus stems for an RNA family with high accuracy via unsupervised clustering of unaligned homologous sequences.