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

In Figures 1-4 , we focus on simulations that were sufficiently converged to produce multipleoccupancy clusters, a hallmark of FARFAR2 sampling being able to discover lowest energy states, as evaluated in Rosetta's all-atom energy function. Other runs with poorer convergence are provided in Supplementary Figures 1-3 ; these cases would need more computer power to discover lowest energy states. In Table 1 , we report the E-gap: the difference in Rosetta energy units (R.E.U.) for best-scoring models in each cluster compared to the top-scoring model in the simulation overall. Rosetta energy functions have been fit such that R.E.U. estimate energies in kcal/mol, 14 so E-gap values similar to or smaller than 3.0 indicate structures that are predicted to make up a significant fraction of the ground state ensemble and that may be trapped by small molecule drugs without a major cost in binding affinity.