Dear all: I've collected a data set for my crystal, which is a two-protein complex. However, normal molecular replacement method gives no solution and thus didn't work. Right now I'm struglling with the phenix.mr_rosetta method. However, I've got several problems when reviewing it, and thus looking forward to get some advice here. 1: There are such words *"NOTE 2: If your structure contains more than one chain or requires more than one homology model to represent the structure, then you need to use mr_model_preparation and phenix.automr to place your model. Then you can start phenix.mr_rosetta with already_placed=True**"* in phenix documentation page. My understanding for this sentences is: the solution derived from *automr* is treated as an original searching model for phenix.mr_rosetta, in which case, we should add the command line *"run_prerefine=True \ number_of_prerefine_models=1000"*. However, what is *mr_model_preparation * for? And if no solution is given by *automr*, what should I do? What's more important, why not endow the phenix.mr_rosetta program to be capable of dealing with data set from multi-chain protein complexes? Since I thought the idea of preliminary manipulation by the program *automr* would decrease the power of *phenix.mr_rosetta*. 2: Two .hhr files are produced. But as displayed in the phenix.mr_rosetta documentation page, only one .hhr file like *"hhr_files=bfr258e.hhr"* are listed in the command line. Why? Could somebody give me any advice on this, please? Thank you and best regards chen -- Cheng Chen, Ph.D. Candidate Laboratory of Structural Biology Life Science Building,Tsinghua University Beijing 100084 China -