Re: [ccp4bb] How large should the real space correlation coefficient be?
Maybe it is worth recalling some ancient discussions, involving Real Space R factors as defined by Alwyn Jones and Gerard Kleywert. If I remember properly, they give an Rfactor between the density in an ATOMMAP generated from a model, but with truncated B factors and the density in the map underconsideration - an exptly phased one, a 2mFO -DFC or whatever. This requires that the electron densities are more or less on the same scale, and gave good Real Space R factors for atoms with low B factors, and high ones for wrong residues, disordered residues, and those with high B factors The CC is meant to avoid problems of scale - the ATOMMAP is calculated taking the b factors into account, so gives a reasonable CC for correctly placed atoms with high b factors. However theoretically a residue with occupancies =0.00 which lies in a totally empty part of the map under consideration could still give a resonable CC . As Pavel says, it is a very blunt tool which can mislead but also help you pinpoint errors.. I have found it most useful when trying to select the best phasing procedure.. Eleanor Pavel Afonine wrote: Hi Hailiang, On 5/25/10 8:14 PM, Hailiang Zhang wrote: Have seen the real-space correlation used widely judging the map quality. Generally or empirically, in order to say an map (area) has good quality, how large should the real space correlation coefficient be? Say, is 0.8 good enough on a residue base? Any references about this will be greatly appreciated! why don't you just familiarize yourself with the map CC values computed per atom or per residue, for a few different structures at different resolutions? It might take you a few hours but from that point on you will have some reference between the map CC values and actual map appearance. phenix.model_vs_data or phenix.real_space_correlation can compute all these values for you. I did it at some point to educate myself and never regretted about the time I spent doing this -:) Pavel.
Re: [ccp4bb] How large should the real space correlation coefficient be?
Hi Hailiang, On 5/25/10 8:14 PM, Hailiang Zhang wrote: Have seen the real-space correlation used widely judging the map quality. Generally or empirically, in order to say an map (area) has good quality, how large should the real space correlation coefficient be? Say, is 0.8 good enough on a residue base? Any references about this will be greatly appreciated! why don't you just familiarize yourself with the map CC values computed per atom or per residue, for a few different structures at different resolutions? It might take you a few hours but from that point on you will have some reference between the map CC values and actual map appearance. phenix.model_vs_data or phenix.real_space_correlation can compute all these values for you. I did it at some point to educate myself and never regretted about the time I spent doing this -:) Pavel.
Re: [ccp4bb] How large should the real space correlation coefficient be?
Hi Pavel: This is actually something I am doing right now. Yes, sometimes it is always better to try it practically. Best Regards, Hailiang Hi Hailiang, On 5/25/10 8:14 PM, Hailiang Zhang wrote: Have seen the real-space correlation used widely judging the map quality. Generally or empirically, in order to say an map (area) has good quality, how large should the real space correlation coefficient be? Say, is 0.8 good enough on a residue base? Any references about this will be greatly appreciated! why don't you just familiarize yourself with the map CC values computed per atom or per residue, for a few different structures at different resolutions? It might take you a few hours but from that point on you will have some reference between the map CC values and actual map appearance. phenix.model_vs_data or phenix.real_space_correlation can compute all these values for you. I did it at some point to educate myself and never regretted about the time I spent doing this -:) Pavel.
[ccp4bb] How large should the real space correlation coefficient be?
Hi, Have seen the real-space correlation used widely judging the map quality. Generally or empirically, in order to say an map (area) has good quality, how large should the real space correlation coefficient be? Say, is 0.8 good enough on a residue base? Any references about this will be greatly appreciated! Thanks! Best Regards, Hailiang
Re: [ccp4bb] How large should the real space correlation coefficient be?
On Tuesday 25 May 2010, Hailiang Zhang wrote: Hi, Have seen the real-space correlation used widely judging the map quality. Generally or empirically, in order to say an map (area) has good quality, how large should the real space correlation coefficient be? I do not think that the real space correlation coefficient is a measure of map quality per se. You could have an excellent experimental map but a lousy model and hence a poor correlation coefficient. Ethan Say, is 0.8 good enough on a residue base? Any references about this will be greatly appreciated! Thanks! Best Regards, Hailiang
