The results I am getting from localized refinement have been showing density that doesn’t correspond to the expected density, as well as artefacts where the mask is placed regardless of mask size.
I have been working on localized refinement of a Hexamer with a molecular weight of 240 kDa. Each of the subunits are about 60 kDa, the molecule has D3 sym, and each subunit bends slightly as you move toward the periphery of the protein. This causes the area, and the active site to be poorly resolved.
I am trying to run localized refinement on a single subunit, so I have used a mask to subtract away the other 5 subunits, and set a fulcrum point in several locations that I think the range of motion can be centered. All of this was along the guidelines for the Yeast U4/U6.U5 tri-snRNP localized refinement, with the exception that I made my mask using segger instead of the volume eraser.
I don’t think the mask placement is wrong because particle subtraction was successful, and no density shows up around my local refinement mask. I have tried increasing the size of the dynamic mask, limiting alignment resolution to 5 angstroms, and increasing the soft mask size, and I am still getting a ripple effect at the mask, and noise outside of that.
Hi Lucas - if I understand right, you are performing signal subtraction in C1, and then focusing on a single subunit of an approximately D3 symmetric homohexamer - is that right?
In this case I would suggest initially not performing signal subtraction, and instead performing symmetry expansion, followed by local refinement with a mask around a single protomer - have you tried that?
@Lucas I think @olibclarke suggestion is a good idea. Can you also check whether you have “enforce non-negativity” enabled in the local refinement jobs, and try without it?
Solved. @olibclarke was correct. defining the center of the mask as the fulcrum and running on a symmetry expanded data set solved the issue. I would also like to add that this problem would still give ripples if the pixel shift and angular rotation parameters were high at all. I used 1 for both shift and rotation to get the best result.
