Poor quality of local refinement

Hi there, I have tried the local refinement for multiple times for a flexible region in my map, but failed to get any improvements. I generated the local mask for the flexible region and a small portion of the neighbor density. Most of the time, after local refinement, it got worse. I am wondering if there is any point that I need to pay extra condition.

Thanks a lot!

1 Like

Hi @Gigi,

Thanks for the post. We have a detailed guide page on the Local Refinement job discussing some of the main considerations with the use of the job, including best practices with regards to masking, parameter settings, etc. We also have a page dedicated to outlining options for mask generation, and finally, a case study on an ideal candidate dataset for local refinement.


1 Like

In my experience, local refinement works really well at taking very similar structures from very similar poses and combining them for very high resolution. It works best on relatively small populations (50k or less particles maybe) and I use it to get the last ~0.5Å improvement of a subregion of an already final region. It doesn’t seem to work particularly well to align particles that are dissimilar or unaligned already quite well. But my region of interest is small, ~30-60kDa so there is not much to align to; i cannot speak for larger domains. I use relatively tight masks (3extend 3soft or so for 1Åpixel), and you might find improvement using “dynamic” instead of static masking. Your particle/region may also be disordered or conformationally heterogeneous, not just positionally unaligned.

1 Like