Xinhua Ji (NCI): Concluding Remarks - During our discussion, I also consulted with the instructors and lecturers of the Cold Spring Harbor Laboratory 2014 course X-ray Method in Structural Biology. Taken together, the following are recommended: (1) Date sets used for structure solution and refinement should be deposited as they are; (2) For the highest resolution (outmost) shell, the I/s(I) value should be  > 1.0 and the completeness should be > 50%; (3) At the claimed resolution of the structure, the overall completeness should be > 93% and the completeness for the outmost shell should be > 70%; and (4) Regardless of the claimed resolution, every reflection in the data set should be included for refinement and map calculations, and the entire data set (without truncation and sigma cutoff) should be deposited with the structure.

*Value to be tested and optimized.

Xinhua Ji (NCI): Click and see my recent lab seminar at the Macromolecular Crystallography Laboratory (MCL) of the NCI for further discussion. Please let me know if you would like to discuss on this topic.

Mariusz Jaskolski (Polish Academy of Sciences): Thanks very much for initiating a discussion about the use of high-resolution reflections for refinement and at other stages of structure determination/publication.  I have a lot of comments and practical remarks in this area, and I have summarized some of them in a one-page document.

A number of similar questions are discussed in my chapter in one of the recent Erice books: M.Jaskolski (2013), High resolution macromolecular crystallography. In: Advancing Methods for Biomolecular Crystallography. R.Read, A.G.Urzhumtsev, V.Y.Lunin eds. Springer, 259-275.

Mark Mayer (NICHD): I understand the benefit of using weak and incomplete data in high resolution shells for calculating maps and improving model building, especially with the routine use of rpim, cc and cc* at the stage of scaling supporting use of reflections in shells with with I/sigma < 2, but I don't understand how to proceed to the deposition/publication stage. After completing model building and refinement using all the data, why would we drop weak and incomplete data in the last round of refinement  to achieve  > 70% completeness and I/sigma > 2 or some other arbitrary cut off that will satisfy reviewers/PDB annotaters? If maps improve with weak and incomplete data in high resolution shells, then there is useful structural information, so why throw it away?

* Files for PDB deposition.