Perhaps these were all mentioned already, but here are some points...looking at your statistics, I would definitely process it to even higher resolution, not lower. I understand the detector might not have been close enough for that (happens often nowadays). You might start seeing hydrogens, double bonds, bent aromatic rings, etc., which I personally think is cool :) and often also informative.
Your problem might be, as mentioned, in the low-resolution data; overloads, a small overlapping satellite crystal, ...? Your pseudotranslation could be higher than the apparent one, since simultaneous twinning will partially hide it (and vice versa; twin tests may tell your crystal is fine, even when it is not). Your a and b cell edges are nearly identical. Did you try refining using that twin operator? What are the processing/refinement stats if you process in the higher symmetry? Since you have high redundancy, did you try to process only the beginning part of the dataset, up to nice completeness and stats? Do you still get the high R factors in refinement? Are your low-resolution R factors so high already at the beginning of data collection? I would expect those to be around 1-2% for this kind of data. Do you have a very low solvent content? This could increase your R factors. And perhaps not the correct forum for it: did you try processing manually with XDS? Having said all that, if your maps are fine, you can show what you see is reliable, and you can explain the high R factor, everyone should be happy :) even reviewer 2. Petri Petri Kursula ---------- Professor Department of Biomedicine University of Bergen, Norway http://www.uib.no/en/rg/petrikursula <http://www.uib.no/en/persons/Petri.Kursula> [email protected] <mailto:[email protected]> ---------- Group Leader, Adjunct Professor Faculty of Biochemistry and Molecular Medicine University of Oulu, Finland ---------- > On 10 Oct 2018, at 12:18, Barone, Matthias <[email protected]> wrote: > > I agree with Pavel... if you got so many observations you dont need TLS but > use anisotropic adps. What I want to add is the dangerous comment to "cutting > back data ... to improve Rfact" Of course, if you lower the amount of > observations while keeping the parameters to fit constant will lower your R > facts. That doesnt necessarily mean you improve your model. It just means > that you incase the param to obs ratio - which says nothing about the > accuracy of your model... > > Dr. Matthias Barone > AG Kuehne, Rational Drug Design > Leibniz-Forschungsinstitut für Molekulare Pharmakologie (FMP) > Robert-Rössle-Strasse 10 > 13125 Berlin > Germany > Phone: +49 (0)30 94793-284 > From: CCP4 bulletin board <[email protected]> on behalf of Pavel Afonine > <[email protected]> > Sent: Wednesday, October 10, 2018 11:03:49 AM > To: [email protected] > Subject: Re: [ccp4bb] Issue with high Rfree (0.25) for a high-resolution > dataset (1.05 Ang) > > Hi Tony, > > I always feel some people are too "greedy" with the resolution they want to > achieve. I mostly find that extremely high density is a pain to work with as > it's usually accompanied by many dual, triple conformers, a lot of noise in > the solvent phase that is often difficult to interpret, etc... Often you > spend more time on a high resolution structure that clearly shows your > protein, as opposed to a low resolution structure where it's difficult to > interpret parts of the map. Also, in most cases you REALLY don't need a 1 Ang > map to clearly show the overall structure of your protein, ligands, first > shell of solvent molecules on the surface of your protein, etc... > > higher resolution typically means you have a chance to obtain a more accurate > atomic model of a crystal structure, which in turn may lead to a more > accurate interpretation of this model. I fully agree that high resolution > requires more modeling effort. There is still a great room for software > developers to provide more automation. > > Your completeness is 90% in your high resolution shell, which is fine, but > have you checked you can clearly see most reflections for h, k and l? Maybe > you're missing many reflections for one of them. I would at least try cutting > your data back to 1.1 or 1.2 Ang, as it might dramatically improve your R > factors and still show everything you want to show. > > Also, did you try TLS refinement? > > At resolutions like 1.2A or better one normally models ADPs as anisotropic > for all atoms (except H), so no need to use TLS. > > Pavel > > > To unsubscribe from the CCP4BB list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1 > <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 > <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
