Hi, Shiyong,

Sure. The RMSD are from the LS superimposed structures. Using the crystal structure of the same protein as the template (perfect alignment here), the backbone RMSD are great (under 0.3) but if all the heavy atoms are selected, the RMSD jumped to around 3.0 angstroms. Here, only 10 models (instead of 50) generated. So, it is an 'initial' result. But 3.0 angstrom is really surprising given the perfect template. Since the only parameter changed is the MD_level, all other default parameters were used. Perhaps the result is what we expect?

Best wishes,

Jian Hui