Thanks Pavel - I'll look forward to playing with that. Another (somewhat speculative) area I think this may be useful is in the combination of high and low-resolution (or low and very low resolution) data sets. Just summarising to check whether I understand correctly:
Say you have a crystal where, say, 50,000 heavy atoms diffract to 2A resolution, but 1,000 atoms in more mobile loops, glycans etc. only have an effective resolution of 4A. The intensity of the reflections around 4A is far and away dominated by the large number of relatively sharply defined atoms, so the weaker/smeared contributions from the mobile atoms are lost in the noise. The result is that in the map, the density comes out looking like it's been cut by a knife, with no clue whatsoever to guide the building of a model for your loop. On the other hand, in a crystal that diffracts only to 3-4A, these loop atoms are on a more even footing with the rest, and their contributions to reflections are non-negligible. In these maps, the density tends to fade out much more gradually, allowing (with care!) the building of a much more complete model. The careful co-refinement of multiple datasets seems like a potentially nice way to get the best of both worlds - and may provide some use for those relatively poorly-diffracting crystals that are often thrown out in favour of those that diffract to higher resolution.
Cheers,
Tristan
Hi all,
It's already common practice with low-resolution crystals to apply torsion-angle restraints on NCS copies and/or to higher-resolution reference structures. But say you have two (or more) crystals of the same protein grown under different conditions, and diffracting to comparable (low) resolution, where no NCS or high-res reference structure is available. In principle, the idea of refining these simultaneously with torsion-angle restraints to each other seems not much different from refinement with NCS - but I'm wondering, has this ever been done?
Best regards,
Tristan