All-at-once RNA folding with 3D motif prediction framed by evolutionary information

Area of Science:

• Structural biology
• Computational biology
• RNA structure prediction

Background:

• Recurrent 3D structural elements in RNA loops, often involving non-Watson-Crick interactions, are crucial for tertiary structure formation.
• Predicting these RNA 3D motifs and their arrangement with secondary structures remains a challenge.

Purpose of the Study:

• To introduce CaCoFold-R3D, a probabilistic grammar for joint prediction of RNA 3D motifs and secondary structures.
• To leverage evolutionary information from RNA alignments for accurate helix and motif identification.

Main Methods:

• CaCoFold-R3D utilizes probabilistic grammars and evolutionary covariation to identify canonical helices and RNA 3D motifs.
• The R3D grammars incorporate helix covariation to constrain the positioning of RNA 3D motifs.
• Predictions are made over a comprehensive set of known RNA motifs, considering their placement in various loop regions.

Main Results:

• CaCoFold-R3D successfully predicts RNA 3D motifs and secondary structures simultaneously.
• The method demonstrates effectiveness in identifying canonical helices, including pseudoknots, using covariation.
• Results validate CaCoFold-R3D as a viable tool for predicting all-residue interactions in RNA 3D structures.

Conclusions:

• CaCoFold-R3D offers a fast and customizable approach for RNA 3D structure prediction.
• The method shows potential for novel motif discovery and can serve as input for deep learning models.
• CaCoFold-R3D can guide RNA design for therapeutic applications, targeting drug development.