Introducing HORNET, a new RNA structure visualization method that correlates sequence and 3D topology

Researchers at the National Cancer Institute have developed a method called HORNET to characterize the 3D topological structure of large, flexible RNA molecules. The scientists used atomic force microscopy (AFM) with deep neural networks and unsupervised machine learning to capture individual conformers under physiological conditions.

Human RNA is transcribed with structural elements important for biological function. Understanding these structures with traditional methods such as cryo-electron microscopy requires highly homogeneous samples and signal averaging. Large, flexible, and heterogeneous RNAs are often difficult to analyze because they adopt multiple conformations when in solution.

No large RNA structure database exists to correlate sequence and 3D topology. Successful protein-centric methods like AlphaFold remain unavailable for RNA, creating a critical gap in structural biology. The general absence of RNA-specific deep learning approaches likely reflects the challenges in obtaining reliable structural models.

In a study published in Nature, “Determining the structure of RNA conformers using AFM and deep neural networks,” scientists introduce HORNET and demonstrate its ability to detect previously hidden large, flexible RNA structural features. Learn more about our innovative features.

The researchers collected single-molecule AFM images of a benchmark RNA in different conformations. They then applied unsupervised machine learning and deep neural networks to correlate molecular topography with energy distribution.

The system was trained on a pseudostructural database covering a wide range of RNA folds and tested on multiple RNAs >200 nucleotides in length (RNase P RNA, cobalamin riboswitch, group II intron, and HIV-1 Rev response element) It was done. RNA). Various initial models were used, including predicted structures and conformers derived from small-angle X-ray scattering data.

In our test case, HORNET accurately reconstructs the three-dimensional structure of individual RNAs, and the root-mean-square deviation (a measure of how well the calculated structure matches the reference) We demonstrated that it is often below the widely used 7 Å threshold for feature confirmation.

Benchmark experiments using simulated and experimental AFM images confirmed the reliability of the combination of previously established constraints and AFM pseudopotentials.

Validation showed that the three-dimensional structures of diverse RNase P RNAs and HIV-1 Rev response element RNAs can be visualized at the single-molecule level. The estimation accuracy from the deep neural network is consistent with the actual distance from known structures.

HORNET addresses critical challenges in RNA structural biology by providing a comprehensive and direct method to interrogate previously elusive RNA structure. This has significant implications for future research across multiple clinical, pharmaceutical, and biotechnological applications.

© 2024 Science X Network