Hi I forgot something - "paired refinement" is the current gold standard for determining a sensible resolution cut-off (see other ccp4bb discussions on the topic).
On 10 Oct 2018, at 11:49, Harry Powell wrote: > Hi > > As a referee, if I had been supplied with data processing statistics that > indicate there is good information going to a higher resolution than that > used for refinement etc, I would want to know why the data had been cut off. > "Making life easier" wouldn't cut the mustard, I'm afraid! > > Clearly, the CC1/2 of 0.796 in the high resolution shell shows that there is > good internal agreement to at least 1.05Å - this internal agreement is > normally taken to mean that the data are "good to that resolution". The mean > I/sd(I) is also quite high in the high resolution shell, which is also > encouraging. I might be inclined to ask why the data were cut to such a low > resolution of 1.05Å anyway... > > A quick back of the envelope calculation suggests that cutting back from 1.05 > to 1.1Å is throwing away ~15% of the data (cutting back to 1.2Å throws away > ~1/3 of the data). Not sure I'd want to do that without a really good > justification. > > However, Michail Isupov raises a very important point; one of the things that > I always teach when discussing "Table 1" is that the statistics for the low > resolution shell should be the best, those for the high resolution shell > should be the worst, and the overall statistics should be in between (but > with luck and a following wind closer to those for the low resolution shell). > Anything else indicates a problem either in the data collection or the data > processing; with very high resolution data, the first things I would look at > are possible overloads in the low resolution shell (in iMosflm there's a > plot of the overloads which is not displayed by default but it's easy to turn > it on...). > > The real answer involves looking at the data processing in detail before > cutting out any data, as others have said. Maybe even taking some time to > look at the diffraction images themselves. > > Overloading high I reflections (which tend to be at low resolution...) is > very easy when chasing high resolution reflections, and is a reason to > collect data in multiple passes to get the best measurements of both strong > and weak reflections; it's something that used to be quite common amongst the > people who collect good datasets. At least some lab-based small molecule > instruments do multiple passes automatically. > > > On 10 Oct 2018, at 11:13, Antonio Ariza wrote: > >> Hi Pavel, >> >> Obviously higher resolution typically means a more accurate atomic model of >> a crystal structure, but I also think that a 1.05 Ang structure is only >> going to provide you with a great deal more insight than a 1.1 or 1.2 Ang >> structure in very specific cases. Particularly if your stats aren't >> completely ideal , why not make sure you get the best possible stats out of >> your data by slightly cutting the resolution, which is still going to leave >> you with a VERY high resolution data set (at least I would consider 1.1 or >> 1.2 Ang as very high resolution)? >> >> Please forgive my ignorance, but I have to admit I haven't researched this >> topic. Is there something wrong with using TLS refinement coupled with >> anisotropic refinement in Refmac? I just checked and I have found that TLS >> still provides a small drop in the R factors when comparing anisotropic >> refinement to TLS + anisotropic refinement in the 1.2 Ang data set I'm >> currently working on. Ok, it's just under 0.5%, which I admit isn't a huge >> change, but shouldn't I take any improvement I can get? >> >> Cheers, >> >> Tony >> >> ------------------------------------------------------ >> >> Dr. Antonio Ariza >> University of Oxford >> Sir William Dunn School of Pathology >> South Parks Road >> Oxford >> OX1 3RE >> e-mail: antonio.ar...@path.ox.ac.uk >> Tel: 00 +44 1865 285655 >> >> Links to my public profiles: >> ResearchGate >> LinkedIn >> GoogleScholar >> Twitter >> >> Check out my latest paper!!! >> Structural insights into the function of ZRANB3 in replication stress >> response >> >> >> >> To unsubscribe from the CCP4BB list, click the following link: >> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1 > ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
