Dear Prof. Powell and Prof. Dodson,
Thanks for your reply and advise.
As per your suggestion, I have re-scaled the intensities using Aimless at 1.861 A.
I observe that the I/sigI has dropped to -0.8 and the behaviour of CC1/2 is still anomalous.
Also, when I inspect the Wilson plot (Fig. 1), I observe that the curve does not fall smoothly with respect to the reference curve (blue). Even with respect to one more Wilson plot from CCP4 website (Fig. 2), the curve from my aimless output is different and discontinuous.
The second moment of I is constant only up to a resolution of 2.4 Å at a value of 3 (Fig. 3). I was not able to get some other plot to compare against mine.
Please tell me if I can still go ahead and refine at 1.861 A.
Thanks,
Satvik
EleanorThirdly - just use all your data and check the refinement results. You can always reject some later if the refinement looks bizarre.First comment - your data looks excelent - even to 1.8A.Second - look also at the wilson plot and second moment plots - these should not be jerky - wilson plot falling off smoothly, 2nd moment pretty constant to the good resolution limit..On 4 August 2017 at 15:37, Harry Powell <[log in to unmask]> wrote:Hi SatvikIf you're only just starting out, the first thing I would recommend is that you forget all about Scala! It's been effectively obsolete for 5-6 years, and any bugs that exist in it will not be fixed.It was replaced by Aimless a long time ago; Aimless is much faster than Scala and the scaling model is somewhat improved (what would you expect with another half-decade of development?).Cutting your data at the data processing stage is probably too early in the structure solution process.In your case, I'd continue analysing your structure with all your data (to 1.861Å,) after re-scaling with Aimless ;-), then once the structure is complete and refined, use the "paired refinement" technique to see if the high resolution data are actually contributing any useful information.HTHHarry--Dr Harry PowellChairman of International Union of Crystallography Commission on Crystallographic ComputingChairman of European Crystallographic Association SIG9 (Crystallographic Computing)
On 4 Aug 2017, at 13:48, Satvik Kumar wrote:<040817_scala.pdf>Dear Crystallographers,
I am a beginner to solving structures by x-ray crystallography. Having collected a diffraction data set using laboratory source X-ray, I am now processing the data. I have integrated the data successfully with the program iMOSFLM using the space group P212121.
During the process of scaling and merging intensities using Scala, I was confronted with the question of setting the resolution cutoff. To clear the doubts I had, I have read the below articles with great care:
1. How good are my data and what is the resolution? Evans and Murshdov, 2013.
2. Assessing and maximizing data quality in macromolecular crystallography. Karplus and Diederichs, 2015.
Based on these articles it becomes very clear that I should set the resolution cutoff to a value at which CC1/2 reduces to 0.2-0.4 irrespective of the values of either Rmerge or I/sigI (prone to uncertainities).
When I ran the program Scala using high resolution cutoffs ranging from 1.861 A (resolution at which data was collected) to 2.33 A, I observe that at 2.0 A, the CC1/2 is 0.582 (outer shell) but the CC1/2 increases to 0.722 at 1.861 A.
The Karplus and Diederichs article clearly states that the CC1/2 is 1 at low resolution and drops down with increasing resolution. It also states that any anomaly should be carefully inspected.
I request the community to kindly inspect the statistics and guide me in setting the optimum resolution cutoff. I have attached a pdf file containing the Scala outputs (statistics and a plot of correlation coefficients versus resolution) for your inspection.
Thanks,
Satvik