Dear Bei, It can sometimes be difficult to tell a real molecular replacement solution from noise at low resolution. In addition to Eleanor's excellent advice, you might try to use your potential derivative data to test the top molrep solutions, even if their statistics are crappy. Try making derivative minus native and anomalous difference maps phased with possible molrep solutions, and see if you get peaks that agree (another nice reality check is to see if those peaks are bigger than the holes). Also, don't despair until you've tried solvent flattening (or more sophisticated solvent massaging). Also remember that your first map will probably be at somewhere around 6A, so beta sheets at best will look like walls and helices will look like tubes (or, as my post doc advisor called the very first map I calculated all by myself, dog turds). Again, don't despair: if you have a map that shows you where the helices are, molecular replacement with those phases, or simple manual placement of the model, should be easy. Good luck, Phoebe
> On Apr 27, 2015, at 7:51 PM, joy yang <[email protected]> wrote: > > Hi All, > > I am looking for some advices on experimental phasing at low resolution, any > advices will be highly appreciated. > > I have 3 data sets, with similar but not identical cell parameters, > Redundancy of each dataset is larger than 6 for overall and 2.5 for > anomalous signal, completeness >95% for overall and >90% for anomalous signal: > > peak: 72.12, 92.07, 128.05, 90, 95.56, 90, P12(1)1, overall Res: 4.5 A, ano > signal 7.13 A (cc>0.15), HA=TaBr, f''=20.99 > > edge: 72.34, 92.28, 127.47, 90, 95.61, 90, P12(1)1, overall Res: 4.1 A, ano > signal 8.31 A (cc>0.15), HA=TaBr , f'=-24.17 > > high remote: 72.18, 92.43, 128.11, 90, 95.60, 90, P12(1)1, overall Res: 4.2 > A, ano signal 12.48 A (cc>0.15), HA=TaBr > > When I tried phenix autosol, the Hyss search gives out a FOM at around 0.30, > which is on the boarder line, however the BayesCC is always between 10-20 > (meaning handiness not distinguishable?), the SKEW is some where between > 0.01-0.05, when it is 0.05, clear solvent boundary could be seen, but the map > is not traceable if it is meaningful at all. > > As I have no idea of how many TaBr cluster could bind, I have tried > everything from 1 to 8, and all I got are similar FOM, SKEW and BAYESCC > values, with the FOM at sites=5 or 6 better than the others. > > I am wondering if anybody have any experience with such kind of data: > 1. what is a better software (maybe SHAPR or SHELEX?) to use in such a case? > > 2. What parameters should I change to make Phenix autosol to work? > > 3. I also have multiple native datasets and peak datasets (TaBr or Pt) which > have quite different cell parameters than these MAD datasets, could these > datasets be of any help? > > 4. What is more, a homolog structure is available, though can not solve the > structure through MR, is there anyway to use the homolog structure as a aid > for experimental phasing? > > Thank you very much in advance! > > Bei >
