If your partial solution suggests dimers or tetramers try searching with those. If you can get  decent looking unit cell packingwith MR, I had good luck using those phases to run parrot and buccaneer with n-fold symmetry to autobuild 90+ % of the final structure. My initial model was about 29% identity. And I was working with twin maps.

__________________
Roger Rowlett
Gordon & Dorothy Kline Professor
Department of Chemistry
Colgate University

On Apr 5, 2016 4:30 PM, "Thomas Krey" <[log in to unmask]> wrote:
Dear all,

I currently work on a 2.6A structure of a ~50kD protein that gave rise to orthorhombic crystals with a large unit cell (95.32  238.58  263.89   90.00   90.00   90.00). According to Matthews there are between 10 and 14 molecules per AU and we only have a replacement model with ~25% aa identity (although it is assumed that it is structurally very conserved). I have already optimized the search model using Phenix ensembler (using two homologous structures).
Phaser is currently trying to explore a very large number of potential solutions with not much signal to distinguish among them and I wonder if there might be a way to 
1) tweak the replacement run in a way that is more straightforward
2) find out, whether this long run is leading anywhere 

The current top solution looks as follows in the log file (after placing two molecules):


Thank you so much for any help or suggestions.

Best wishes
Thomas