Hi Alexei,
   In case of very good main chain density like you have, you can decide on N->C directionality by looking for the slant of the C=O group in helices. As you go from N->C, the C=O has the tendency to slant upward. In case of beta sheets, it may be difficult to judge, but at this point, it may not be that crucial. You can build whatever strands/sheet fragments you wish as polyALA and the use your partial model to get more phase information. You should be able to sort out your directionality later on.
 
Another thing to aid in getting N/C term orientation is to try to do secondary structure prediction on your protein alongwith fold prediction/threading methods in case your sequence is very unique. Comparison of secondary structure prediction with your experimental electron density maps should provide additional valuable clues about orientation.
 
I had a case recently with low res phases where we were able to trace a partial starting model, which was then used in a DALI search to obtain a better structural model for completing the structure determination.
 
You plan for phasing strategy sounds fine, you might want to include search for additional heavy atoms by difference Fourier methods using your partial model phases in case you have not found all possible sites. Getting additional sites may help your starting experimental phases together with combination of partial model phases with exisiting phases.
 
Hope this help.
 
-Debanu.
--
Debanu Das,
JCSG,
SSRL-SLAC.

________________________________

From: CCP4 bulletin board on behalf of Alexei Datsuk
Sent: Fri 6/15/2007 10:03 AM
To: [log in to unmask]
Subject: [ccp4bb] Low resolution chain tracing


Hi,

I need some advice from experts.  I have a low resolution electron density map (3.7 angstroms) solved by SIRAS.  The protein is a tetramer (47 kDa/monomer) but I'm having trouble getting in register.  There is very good main chain density and clear secondary structure (beta-sheets) so I likely have solved it correctly.  I can trace most of the main chain as fragments (loops connecting beta strands are broken) but I can't tell if I'm tracing it in the right direction or whether I have it backwards.  There is no real density for side chains at all (except for Calpha), not even aromatics, so I have no clue which way is N or C-terminus.  There are a few disulfide bonds too so I'm real worried about tracing incorrectly and never being able to fix it later.  My strategy right now is to build in as much poly-ala fragments (I can get ~75% of trace as fragments) and then refine them.  Then redo phasing in SHARP with poly-ala model and SIRAS phases and then use DM with NCS averaging to get
improved map.  Hopefully I can see then side chain density.  My question is how much bias would be incorporated if some of the fragments are traced backwards?  Would I be able to sort things out if some fragments are traced backwards?  Has anyone traced chains as fragments not knowing which direction is which and then be able to sort things out later?  Right now I am stuck- I have spent the last 1 months trying to get things in register but no luck.  I don't want to build haphazardly but it doesn't look like I have much choice.  

What have people done when tracing in low resolution maps?

Thanks

Alexei