I use the low resolution protein docking these days. In one case,I use full atoms during protocol to generate 10000 poses and top 200 best score poses were clustered by cluster protocol. In another case, I just generate the 10000 pose with protein backbone and also cluster the top 200 best score poses. I found that the results between these two cases are different.
I am wondering which one should I use for the final protein docking solution?
If you have no other data whatsoever, then you’re forced to go with lowest score = best result. If you have any experimental hints, you can also use those to figure out which model is most likely to be accurate.