De novo 3D models of SARS-CoV-2 RNA elements and small-molecule-binding RNAs to guide drug discovery

Ramya Rangan; Andrew M. Watkins; Wipapat Kladwang; Rhiju Das

For systems larger than 50 nucleotides, we generated large model sets and clustered the 400 lowest energy structures with a 5 Ã… RMSD clustering radius, the procedure used for similarly sized systems in our recent FARFAR2 modeling benchmark. (Here and below, RMSD was computed as the allheavy-atom RMSD between two models, as in all recent Rosetta work on RNA modeling.) For smaller systems, we clustered the 400 lowest energy structures with a 2 Ã… RMSD clustering radius, analogous to the procedure used for similarly sized systems in the FARFAR2 study. We note here that the clustering method follows the protocol used in the FARFAR2 and other Rosetta RNA studies, where cluster centers are drawn from the very lowest energy models observed (see Methods). We make available models across up to 10 clusters in the resulting FARFAR2-SARS-CoV-2 data set to help efforts in virtual screening that take advantage of ensembles.