With respect to RNA design, rational engineering has yielded versatile sensors and nano-structures 10 - 12, but has so far been limited to rearrangements of existing sequence modules rather than designing new non-canonical structures.

Recent, automated full-atom methods (iFold3D 8, MC-SYM 9) have described models of impressive quality, but non-canonical regions appear to be either incorrect 8 or take advantage of sequence similarity with homologs of known structure within the method's training database 9. Existing de novo methods for modeling tertiary structure have largely been limited to low resolution (e.g., Fragment Assembly of RNA (FARNA) 5, DMD 6) or have required manual atom-level manipulation by expert users (e.g, MANIP 7).

3, 4 A critical requirement for a high-resolution RNA modeling method is its ability to find native-like solutions for the ‘jigsaw puzzles’ presented by these non-canonical motifs.ĭespite their small size, these motifs are often quite complex, with intricate meshes of non-Watson-Crick hydrogen bonds and irregular backbone conformations. The junctions, loops, and contacts that underlie these tertiary structures are frequently less than ten nucleotides in length and, in some cases, are able to self-assemble into the same microstructures when grafted into other helical contexts. 2 A central, unsolved challenge at present is to model how the resulting canonical double helices are positioned into specific tertiary structures. Methods for inferring an RNA's pattern of canonical base pairs (secondary structure) have been well-calibrated and widely used for decades, often in concert with phylogenetic covariation analysis and structure mapping experiments. RNA is an ancient component of all living systems, whose catalytic prowess, biological importance, and ability to form complex folds have come to prominence in recent years 1.

Microsoft jigsaw puzzles accuracy chain