Thank you very much for the help from all of you. I used "no prior phase information", and got R/Rfree = 0.2202/0.2544. I used "SAD data directly" and got R/Rfree = 0.2066/0.2302. I used "Hendrickson-Lattman coefficients" and got R/Rfree = 0.2255/0.2539.
It seems that the "SAD data directly" gets a better result for my SAD data. Best wishes, Q. Cai Qixu Cai Email: caiq...@gmail.com School of Life Sciences, Xiamen University, Fujian, China 2012/11/5 Navraj Pannu <navraj.singh.pa...@gmail.com> > > >> I am not sure how you would fare with SAD data directly.. >> >> > If you would like a comparison, check out the refmac paper. > > http://journals.iucr.org/d/issues/2011/04/00/ba5152/index.html > > See Figure 1 - this shows model building with ARP/wARP on over 100 data > sets using > the different Refmac functions that you asked about. > > The "MLHL" function is using "phase and fom" or "Hendrickson-Lattman" > coefficients > The "Rice" function is "no prior phase information" and the "SAD" function > is using "SAD > data directly". > > The figure shows fraction of the model built automatically. If you see > lot of red squares on the top left of the diagonal (which you do), this > just means the SAD function outperforms the 'Rice' function and you see > more Red squares then Blue circles, so SAD performs better than MLHL. You > can get more information on individual functions from references cited in > the Refmac paper. > > The top right shows data sets between 80-100% built by all functions: > thus, if your model/map is good enough, any function can build it. > > For SAD data sets that we have collected here, we always use the SAD > function even at the end of the refinement and found (unsurprisingly) a > lower difference between R-free and R then other functions. > > As Garib mentioned, Pavol Skubak and/or I are happy to answer any > questions. > > Best wishes, > Raj > > > >> But beware - do the HLs etc combine phases from the existing model and >> the anomalous signal - if so they may be biased towards the existing model, >> and make it harder to correct >> Eleanor >> >> On 5 Nov 2012, at 09:46, Qixu Cai wrote: >> >> > Dear all, >> > >> > What's the difference between "no prior phase information", "phase and >> FOM", Hendrickson-Lattman coefficients", and "SAD data directly" in the >> refmac5 GUI of CCP4i? >> > >> > I have a SAD dataset and solve the phase by phenix.autosol. Now I want >> to refine the structure by refmac5, which kind of input above I should >> choose? >> > >> > Another question is in the "Refinement Parameters" of refmac5. What's >> the meaning of "Use experimental sigmas to weight Xray terms"? If I use >> molecular replacement to solve the structure, shall I "uncheck" this item? >> > >> > Thanks very much for your help. >> > >> > Best wishes, >> > >> > Q. Cai >> > >> > >
