First choose the molecule you mask as the one with the best and most 
complete density and give all the matrices relative to that copy.
Anything which is masked out will certainly not return..

You only need one mask in one of the crystals; the matrices will 
transform that into the otjer crystal form too, providing you specify 
them correctly!
 Eleanor

Jianghai Zhu wrote:
> Thanks for the suggestions.  However, in two different space groups, I 
> know the NCS contacts are similar, but the symmetry-related contact 
> can be different.  Should I generate different masks for different 
> crystals? or it is not a big deal.
>
> Another question.   In my model, there is one domain which is fixed in 
> one molecule (good density, lots of lattice contacts) but flexible in 
> all other 5 molecules (no density, lacking lattice contacts).  When I 
> left this domain out of the mask, the density of the domain was gone 
> after dmmulti.  What should I do with this particular domain?
>
> Thanks.
>
> Jianghai
>
>
> On Feb 22, 2007, at 4:59 AM, Kevin Cowtan wrote:
>
>> Two things:
>>
>> 1. You can do overlap removal in NCSMASK using the OVERLAP keyword.
>>
>> 2. You shouldn't be using 6 masks anyway - you only need one mask. 
>> Dmmulti generates the rest internally.
>
>
> On Feb 22, 2007, at 4:54 AM, Eleanor Dodson wrote:
>> You just need one mask for your master molecule, which will be 
>> transferred by the matrices you give to all the other copies.
>> I make the mask from the crude model, with a generous border round it 
>> - the program will also prune it to get rid of overlaps..
>>
>> Eleanor
>
>
>>
>> Jianghai Zhu wrote:
>>> Dear all,
>>> I have two low resolution (3.8 A) MAD data sets from two different  
>>> space groups.  There are 4 copies of my molecule in one space group  
>>> and 2 copies in the other.  The density-modified maps from these 
>>> data  sets are poor, but still allow me to build a crude model on 
>>> them.   All the domains should be at the right or close to right 
>>> orientations  and positions.   I would like to try multi-crystal 
>>> averaging to  improve my maps.  Since I already have the model, I 
>>> used NCSMASK to  make 6 masks for the 6 copies of my model.  Then I 
>>> used DMMULTI to  perform the multi-crystal averaging.  The maps came 
>>> out of DMMULTI  showed great improvement.  Some densities were not 
>>> seen before showed  up nicely.  However, after I read the manual 
>>> carefully, I realize  that DMMULTI would not take care of the mask 
>>> overlap.  So some  regions in my density map between the NCS-related 
>>> and symmetry- related molecules must be messed up.   Could any 
>>> expert out there  give me some suggestions about how to make proper 
>>> NCS masks for  DMMULTI?  Thanks at advance.
>>> Jianghai
>