MSMs

MSMs were constructed to provide kinetics and free energy estimates. The MSM was built using the PyEMMA v2.5.7 program.⁶⁸ It was not possible to build an MSM using just the features of the 24 dissimilar residues (12 in each protomer) between SARS-CoV2 and SARS-CoV. Therefore, all backbone dihedral angles were selected. In addition, the first χ angle (χ1) from 24 dissimilar residues was also included in MSM building. For MERS-CoV, the χ1 angles from residues at the equivalent position were also selected. Time-lagged independent component analysis (tICA) was used to reduce the dimensionality of the data.^69,70 It was possible to build models that were Markovian with a lag time of ≥10, with the lag time selected according to the convergence of the implied timescales. The dimension reduction was achieved by projecting on the three slowest tICA components. The K-means clustering algorithm was used to obtain 100 microstates. The conformational clusters were grouped together based on the kinetic similarity using the PCCA+ algorithm.⁷¹ The PCCA+ algorithm uses the eigenvectors of the MSMs to group together clusters, which are kinetically close, resulting in a set of macrostates. The final number of metastable macrostates was selected based on the implied timescale plot. The MSMs were validated using the Chapman–Kolmogorov test implemented in PyEMMA.⁶⁸