On 17/09/18 19:49, Robert Grant wrote: > I have been using the Phenix-linked Coot to modify a MR solution calculated > from a rather poor data set, with high merging Rs and very low absolute > intensities. I have observed what I consider bizarre behavior in the real > space refinement function. My original observation was that very often RSR > would actually make the model look less like the density. I generally use > the default matrix weight (60) or reduce it to 20. When I let coot estimate > the weight it came up with 65446.
That seems like a high number - and maybe coincidentally close to 65536. > With that value the RSR seems to behave normally. How is this weight > calculated? What kind of data pathology would produce the results I am > seeing? Is it related to the low intensity values in the data? > The "Estimate" button give a value that is proportional to the inverse of the rmsd of the map (so, yes) - if you don't have an average sized protein at average resolution with structure factors for the map more or less on the absolute scale, then your initial estimate will be a bit off. (after having selected a reasonably well-fitting part of the structure) you can also use Extensions -> Settings -> Autoweight Refinement. In my hands refinement is not very sensitive to the weight - as long as you are within a factor of ~5, it seems OK. Paul. ######################################################################## To unsubscribe from the COOT list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=COOT&A=1
