What column labels does phenix.refine tell you it is using ? It's likely that phenix.refine is grabbing your anomalous data I(+)/I(-) or F(+)/F(-) and using that for refinement if you have these types of columns in your MTZ file. This may or may not explain your issue. Completeness of each column label is shown in CCP4's MTZDUMP output for the MTZ file. Use: mtzdump hklin mydata.mtz and then type "go". It's a lot easier to give phenix.refine less latitude to do something unexpected by giving it only the columns you want to use for refinement. Use CCP4's CAD program for that (the LABOUT command to output only the required subset of the data). Obviously this can also provide you with the answer to your question relating completeness in refinement and data column usage. Phil Jeffrey Princeton On 12/4/12 5:47 PM, Yarrow Madrona wrote:

Hi, (Phenix v.1.8.1-1168)

I recently refined a structure using experimental Intensities and compared it to the same structure refined with calculated structure factors.

I know if you provide intensities Phenix.refine will convert to structure factors. I am not sure if it runs truncate or what. Anyway, I get this in the log file which matches my completeness in the scala log file:

Completeness in resolution range: 0.992202 Completeness with d_max=infinity: 0.992164

When I run using structure factors (Truncate run in Scala) I get the following:

Completeness in resolution range: 0.901282 Completeness with d_max=infinity: 0.901259

Perhaps Scala did not merge anomolous and non-anomolous data? This does not may a whole lot of sense to me. Is some data thrown out? Can someone explain?

Thank you.

On Tue, Dec 4, 2012 at 2:47 PM, Yarrow Madrona wrote:

I recently refined a structure using experimental Intensities and compared it to the same structure refined with calculated structure factors.

I know if you provide intensities Phenix.refine will convert to structure factors. I am not sure if it runs truncate or what.

It runs the French & Wilson treatment that Jeff implemented in CCTBX - this should effectively be the same as what Truncate does.

When I run using structure factors (Truncate run in Scala) I get the following:

Completeness in resolution range: 0.901282 Completeness with d_max=infinity: 0.901259

Perhaps Scala did not merge anomolous and non-anomolous data? This does not may a whole lot of sense to me. Is some data thrown out? Can someone explain?

I don't know the explanation, but phenix.refine does perform outlier rejection, and it's possible that it behaves differently depending on what input you give it. That's an awfully big drop in completeness, though. Can you please send us the logfiles for the two refinement runs off-list? ([email protected] will cover most of us.) -Nat

Hi,

I recently refined a structure using experimental Intensities and compared it to the same structure refined with calculated structure factors.

I know if you provide intensities Phenix.refine will convert to structure factors. I am not sure if it runs truncate or what.

It uses French & Wilson method to convert Iobs to Fobs, which is implemented in Phenix based on original publication.

Anyway, I get this in the log file which matches my completeness in the scala log file:

Completeness in resolution range: 0.992202 Completeness with d_max=infinity: 0.992164

When I run using structure factors (Truncate run in Scala) I get the following:

Completeness in resolution range: 0.901282 Completeness with d_max=infinity: 0.901259

If it was the other way around I could speculate about the reason.. Otherwise I need to see the data file. Could you send it to me off list? Pavel