The range of BSC number for FrealignX

Submitted by AlanZhang on Mon, 2017-10-02 16:59

Forums

Frealign

I am wondering the appropriate range for the BSC number. Does it have to be within 0-20? Or we can use any number which could make the FSC curve best, even it is larger than 20?

niko Tue, 2017-10-03 01:15

You can use even higher numbers but please make sure when you see an improved FSC curve that you also see improved density features. The FSC is not always a reliable resolution measure.

AlanZhang Tue, 2017-10-03 15:00

Thank you, Niko.

Sometimes I would get better FSC and higher reported resolution with a larger BSC number especially for a large dataset (more than 500k particles, so I think it is possible that I have more junk particles in these cases). However, the map does not improve too much. The same thing happens when I set a higher res_high_refinement number, in this case the map even gets stretched on the edge (is that over-fitting?). So the question I have is how to tell what is the real resolution. Like if we improve the FSC from 3.3A to 3.1A at the 0.143 criteria by playing with the parameters in frealign but we can not tell the differences between the two maps. Does that mean 3.3A is underestimated or 3.1A is overestimated (it is really hard for me to tell the real resolution just based on the map at these range)? As you might understand we usually like to report the higher number.

AlanZhang Tue, 2017-10-03 15:40

I would expect some ambiguous regions like the loop regions or some areas where the side chain densities are not so clear would become better despite I do not know whether those are supposed to be or not. Anyway, if there is no difference, should I just report the lower number (3.3A but not 3.1A)?