I have a question for you. I have generated models for a sequence using one structural homologue. In one set i have used 1 crystal structure of the reference protein and generated 10 models. In the second set i have used two crystal structures of the same reference protein and generated 10 models.The sequence alignment is the same in both cases. My question is, can i compare the pdf violations of a model generated in set 1 with a model generated in set 2 ?