Have you tried fixing the molecule that looks correct and searching for others? You might have greater than one but less than 9 molecules per ASU.

When you do this, try imposing severe restraints on the packing function. This worked for me in Phaser with a difficult case. My anecdotal experience is that, when you have lots of molecules per asu, the correct solution gets swamped by poorly-packed solutions if the default packing penalties are used.

Good luck.

Sent from my iPhone

> On May 15, 2014, at 6:50 PM, "Matthew Bratkowski" <[log in to unmask]> wrote:
> 
> Hello all,
> 
> 
> I am working on the structure of a small protein in space group P212121.  The protein is monomeric in solution based on gel filtration analysis.  The Matthews Coefficeint program indicates that 9-10 molecules per asymmetric unit results in ~50% solvent content, while 1 molecule per asymmetric unit results in ~95% solvent. 
> 
>  I tried molecular replacement with a search model which is essentially identical in sequence to my protein, and searched for 9 or 10 molecules/asu.  Using MolRep with 9 or 10 molecules/asu, I get poor contrast scores around 1-1.5.  However, when using Phaser, I get a solution with one molecules/asu.  Likewise, when I went back and tried MolRep with 1 molecule/asu, I got a contrast score of 3.12.  This model still has some issues, but looks more correct compaired to models created with 9 or 10  molecules/asu. 
> 
> It seems highly unlikely that a crystal would contain 95% solvent, but is there any possiblility that this could be the case?  Assuming that the Matthews coefficient is correct, does anyone have an idea why MR seems to work better for 1 molecule/asu with 95% solvent content compared to 9-10 molecules with 50% solvent content? Alternatively, is there any reason why the Matthews coefficient could be calculating incorrectly?  Any suggestions would be helpful.
> 
> Thanks,
> Matt