Dear All,   Following is a table of x-ray data analysis. In it the "Outershell" is same as "the highest resolution shell" in the table 1 of crystallography paper, am I right?   I am looking forward to getting a reply from you.   Cheers,   Dialing   Overall InnerShell OuterShell Low resolution limit 77.381 77.381 1.669 High resolution limit 1.663 7.718 1.663 Rmerge 0.059 0.031 0.586 Ranom 0.057 0.028 0.561 Rmeas (within I+/I-) 0.062 0.031 0.625 Rmeas (all I+ & I-) 0.062 0.032 0.620 Rpim (within I+/I-) 0.023 0.012 0.271 Rpim (all I+ & I-) 0.017 0.010 0.196 Total number of observations 171554 1803 810 Total number unique 13699 181 88 Mean(I)/sd(I) 32.4 46.5 4.6 Completeness 98.0 100.0 70.4 Multiplicity 12.5 10.0 9.2  

On 29 Apr 2012, at 23:55, Dialing Pretty wrote:

In some table 1 of the crystallography paper, there is an item called "FOM". Will you please explain the meaning of it? Do we have a server to get it or some other route to get this value?

I'm guessing this is the "figure of merit", defined as the cosine of the phase error. Since the correct phase is generally unknown it's an estimate, which may be calculated during density modification (phase improvement). // Best wishes; Johan

In some table 1 of the crystallography paper, there is an item called "FOM". Will you please explain the meaning of it? Do we have a server to get it or some other route to get this value?

This is the letter "m" in Randy's 2mFo-DFc map -;) phenix.refine, phenix.maps and other relevant tools estimate it in resolution shells using test reflections as described here: J. Appl. Cryst. (1996). 29, 741-744 and here Acta Cryst. (1995). A51, 880-887 Of course there are plenty of other literature on this subject! Pavel

IMO reporting statistics in "Outer shell" or in "the highest resolution shell" doesn't really make sense for obvious reasons which don't need to be explained unless one wants to summarize a crystallography text book in an email. If you want to be minimalistic just state the overall figure. A way better idea, though, is to report completeness (and other similar metrics) in relatively thin resolution bins, which would actually tell something about your data. Pavel On 4/30/12 8:26 AM, Nathaniel Echols wrote:

Correct, assuming that the high resolution used in data processing is the same as the high resolution used for refinement.

-Nat

On Mon, Apr 30, 2012 at 1:22 AM, Dialing Pretty wrote:

...

Dear All,

Following is a table of x-ray data analysis. In it the "Outershell" is same as "the highest resolution shell" in the table 1 of crystallography paper, am I right?

I am looking forward to getting a reply from you.

Cheers,

Dialing

Overall InnerShell OuterShell Low resolution limit 77.381 77.381 1.669 High resolution limit 1.663 7.718 1.663

Rmerge 0.059 0.031 0.586 Ranom 0.057 0.028 0.561 Rmeas (within I+/I-) 0.062 0.031 0.625 Rmeas (all I+& I-) 0.062 0.032 0.620 Rpim (within I+/I-) 0.023 0.012 0.271 Rpim (all I+& I-) 0.017 0.010 0.196 Total number of observations 171554 1803 810 Total number unique 13699 181 88 Mean(I)/sd(I) 32.4 46.5 4.6 Completeness 98.0 100.0 70.4 Multiplicity 12.5 10.0 9.2

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Sure, but then at least one needs to clearly define what exactly is "Outer shell" and "highest resolution shell". I can come up with gazillions ways of defining the "highest resolution shell" and depending on how I define it the numbers will be vastly different. In general it may be a good idea to define what exactly you want to calculate first before cranking the machine to get some numbers. In this sense reporting statistics in resolution bins contains both - the definition and desired numbers. Pavel On 4/30/12 1:58 PM, Nathaniel Echols wrote:

...

IMO reporting statistics in "Outer shell" or in "the highest resolution shell" doesn't really make sense for obvious reasons which don't need to be explained unless one wants to summarize a crystallography text book in an email. I disagree - for the data processing it is very relevant to know what

On Mon, Apr 30, 2012 at 1:49 PM, Pavel Afonine wrote: these statistics are, otherwise one has very little idea what the criteria used to determine "resolution" were. For refinement it is perhaps less essential; the R-factors in the outer shell will almost always be somewhat higher than the values for the entire dataset, but this rarely tells you anything.

-Nat

The problem is the journals sometimes specify the content of Table 1. There might be more room for this if It is supplemental table 1. I bet if you make it a GUI feature (table 1 vs extended supplemental table 1 option) it will catch on fast. Kendall On Apr 30, 2012, at 5:17 PM, Pavel Afonine wrote:

Sure, but then at least one needs to clearly define what exactly is "Outer shell" and "highest resolution shell". I can come up with gazillions ways of defining the "highest resolution shell" and depending on how I define it the numbers will be vastly different. In general it may be a good idea to define what exactly you want to calculate first before cranking the machine to get some numbers. In this sense reporting statistics in resolution bins contains both - the definition and desired numbers.

Pavel

On 4/30/12 1:58 PM, Nathaniel Echols wrote:

...

...

IMO reporting statistics in "Outer shell" or in "the highest resolution shell" doesn't really make sense for obvious reasons which don't need to be explained unless one wants to summarize a crystallography text book in an email. I disagree - for the data processing it is very relevant to know what

On Mon, Apr 30, 2012 at 1:49 PM, Pavel Afonine wrote: these statistics are, otherwise one has very little idea what the criteria used to determine "resolution" were. For refinement it is perhaps less essential; the R-factors in the outer shell will almost always be somewhat higher than the values for the entire dataset, but this rarely tells you anything.

-Nat

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

On Mon, Apr 30, 2012 at 4:22 AM, Phil Evans wrote:

Are anisotropic cutoff desirable?

is there a peer-reviewed publication - perhaps from Acta Crystallographica - which describes precisely why scaling or refinement programs are inadequate to ameliorate the problem of anisotropy, and argues why the method applied in Strong, et. al. 2006 satisfies this need? -Bryan

On Tue, May 1, 2012 at 10:34 AM, Kendall Nettles wrote:

I have seen dramatic improvements in maps and behavior during refinement following use of the UCLA anisotropy server in two different cases. For one of them the Rfree went from 33% to 28%. I don't think it would have been publishable otherwise.

Was this just performing anisotropy correction, or also truncation? It would be interesting to see the data before and after treatment, in any case. I wonder if combining the anisotropy correction with B-factor sharpening of the map coefficients would make a difference. -Nat

Hi Kendall, A caution on using the free R value to evaluate the utility of applying an anisotropic truncation to the data: As you remove data with low I/sigma, you will generally improve the free R regardless of whether your model is improved or not. This is of course true also if you simply remove all the data with I/sigma < 3 or apply any other truncation of that type. All the best, Tom T ________________________________________ From: [email protected] [[email protected]] on behalf of Kendall Nettles [[email protected]] Sent: Tuesday, May 01, 2012 11:34 AM To: PHENIX user mailing list Subject: Re: [phenixbb] Geometry Restraints - Anisotropic truncation I have seen dramatic improvements in maps and behavior during refinement following use of the UCLA anisotropy server in two different cases. For one of them the Rfree went from 33% to 28%. I don't think it would have been publishable otherwise. Kendall On May 1, 2012, at 11:10 AM, Bryan Lepore wrote:

On Mon, Apr 30, 2012 at 4:22 AM, Phil Evans wrote:

...

Are anisotropic cutoff desirable?

is there a peer-reviewed publication - perhaps from Acta Crystallographica - which describes precisely why scaling or refinement programs are inadequate to ameliorate the problem of anisotropy, and argues why the method applied in Strong, et. al. 2006 satisfies this need?

-Bryan _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

I didnt think the structure was publishable with Rfree of 33% because I was expecting the reviewers to complain. We have tested a number of data sets on the UCLA server and it usually doesn't make much difference. I wouldn't expect truncation alone to change Rfree by 5%, and it usually doesn't. The two times I have seen dramatic impacts on the maps ( and Rfree ), the highly anisotrophic sets showed strong waves of difference density as well, which was fixed by throwing out the noise. We have moved to using loose data cutoffs for most structures, but I do think anisotropic truncation can be helpful in rare cases. Kendall On May 1, 2012, at 3:07 PM, "Dale Tronrud" wrote:

While philosophically I see no difference between a spherical resolution cutoff and an elliptical one, a drop in the free R can't be the justification for the switch. A model cannot be made more "publishable" simply by discarding data.

We have a whole bunch of empirical guides for judging the quality of this and that in our field. We determine the resolution limit of a data set (and imposing a "limit" is another empirical choice made) based on Rmrg, or Rmes, or Rpim getting too big or I/sigI getting too small and there is no agreement on how "too big/small" is too "too big/small".

We then have other empirical guides for judging the quality of the models we produce (e.g. Rwork, Rfree, rmsds of various sorts). Most people seem to recognize that the these criteria need to be applied differently for different resolutions. A lower resolution model is allowed a higher Rfree, for example.

Isn't is also true that a model refined to data with a cutoff of I/sigI of 1 would be expected to have a free R higher than a model refined to data with a cutoff of 2? Surely we cannot say that the decrease in free R that results from changing the cutoff criteria from 1 to 2 reflects an improved model. It is the same model after all.

Sometimes this shifting application of empirical criteria enhances the adoption of new technology. Certainly the TLS parametrization of atomic motion has been widely accepted because it results in lower working and free Rs. I've seen it knock 3 to 5 percent off, and while that certainly means that the model fits the data better, I'm not sure that the quality of the hydrogen bond distances, van der Waals distances, or maps are any better. The latter details are what I really look for in a model.

On the other hand, there has been good evidence through the years that there is useful information in the data beyond an I/sigI of 2 or an Rmeas > 100% but getting people to use this data has been a hard slog. The reason for this reluctance is that the R values of the resulting models are higher. Of course they are higher! That does not mean the models are of poorer quality, only that data with lower signal/noise has been used that was discarded in the models you used to develop your "gut feeling" for the meaning of R.

When you change your criteria for selecting data you have to discard your old notions about the acceptable values of empirical quality measures. You either have to normalize your measure, as Phil Jeffrey recommends, by ensuring that you calculate your R's with the same reflections, or by making objective measures of map quality.

Dale Tronrud

P.S. It is entirely possible that refining a model to a very optimistic resolution cutoff and calculating the map to a lower resolution might be better than throwing out the data altogether.

On 5/1/2012 10:34 AM, Kendall Nettles wrote:

...

I have seen dramatic improvements in maps and behavior during refinement following use of the UCLA anisotropy server in two different cases. For one of them the Rfree went from 33% to 28%. I don't think it would have been publishable otherwise. Kendall

On May 1, 2012, at 11:10 AM, Bryan Lepore wrote:

...

On Mon, Apr 30, 2012 at 4:22 AM, Phil Evans wrote:

...

Are anisotropic cutoff desirable?

is there a peer-reviewed publication - perhaps from Acta Crystallographica - which describes precisely why scaling or refinement programs are inadequate to ameliorate the problem of anisotropy, and argues why the method applied in Strong, et. al. 2006 satisfies this need?

-Bryan _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

---

phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Is the explanation not simpler? The volumes of reciprocal space that were left out did not in fact contain signal, and it's by removing those noise non-reflections that the actual R is revealed. As James Holton posted a while ago, Rfactors calculated for noise give randomly large values. So it seems less less misleading to refer to it as "anisotropy-TRUNCATED". phx. On 03/05/2012 07:40, Pavel Afonine wrote:

Hi Kendall,

I just did this quick test: calculated R-factors using original and anisotropy-corrected Mike Sawaya's data (*)

Original: r_work : 0.3026 r_free : 0.3591 number of reflections: 26944

Truncated: r_work : 0.2640 r_free : 0.3178 number of reflections: 18176

The difference in R-factors is not too surprising given how many reflections was removed (about 33%).

Pavel

(*) Note, the data available in PDB is anisotropy corrected. The original data set was kindly provided to me by the author.

On 5/2/12 5:25 AM, Kendall Nettles wrote:

...

I didnt think the structure was publishable with Rfree of 33% because I was expecting the reviewers to complain.

We have tested a number of data sets on the UCLA server and it usually doesn't make much difference. I wouldn't expect truncation alone to change Rfree by 5%, and it usually doesn't. The two times I have seen dramatic impacts on the maps ( and Rfree ), the highly anisotrophic sets showed strong waves of difference density as well, which was fixed by throwing out the noise. We have moved to using loose data cutoffs for most structures, but I do think anisotropic truncation can be helpful in rare cases.

Kendall

On May 1, 2012, at 3:07 PM, "Dale Tronrud" wrote:

...

While philosophically I see no difference between a spherical resolution cutoff and an elliptical one, a drop in the free R can't be the justification for the switch. A model cannot be made more "publishable" simply by discarding data.

We have a whole bunch of empirical guides for judging the quality of this and that in our field. We determine the resolution limit of a data set (and imposing a "limit" is another empirical choice made) based on Rmrg, or Rmes, or Rpim getting too big or I/sigI getting too small and there is no agreement on how "too big/small" is too "too big/small".

We then have other empirical guides for judging the quality of the models we produce (e.g. Rwork, Rfree, rmsds of various sorts). Most people seem to recognize that the these criteria need to be applied differently for different resolutions. A lower resolution model is allowed a higher Rfree, for example.

Isn't is also true that a model refined to data with a cutoff of I/sigI of 1 would be expected to have a free R higher than a model refined to data with a cutoff of 2? Surely we cannot say that the decrease in free R that results from changing the cutoff criteria from 1 to 2 reflects an improved model. It is the same model after all.

Sometimes this shifting application of empirical criteria enhances the adoption of new technology. Certainly the TLS parametrization of atomic motion has been widely accepted because it results in lower working and free Rs. I've seen it knock 3 to 5 percent off, and while that certainly means that the model fits the data better, I'm not sure that the quality of the hydrogen bond distances, van der Waals distances, or maps are any better. The latter details are what I really look for in a model.

On the other hand, there has been good evidence through the years that there is useful information in the data beyond an I/sigI of 2 or an Rmeas> 100% but getting people to use this data has been a hard slog. The reason for this reluctance is that the R values of the resulting models are higher. Of course they are higher! That does not mean the models are of poorer quality, only that data with lower signal/noise has been used that was discarded in the models you used to develop your "gut feeling" for the meaning of R.

When you change your criteria for selecting data you have to discard your old notions about the acceptable values of empirical quality measures. You either have to normalize your measure, as Phil Jeffrey recommends, by ensuring that you calculate your R's with the same reflections, or by making objective measures of map quality.

Dale Tronrud

P.S. It is entirely possible that refining a model to a very optimistic resolution cutoff and calculating the map to a lower resolution might be better than throwing out the data altogether.

On 5/1/2012 10:34 AM, Kendall Nettles wrote:

...

I have seen dramatic improvements in maps and behavior during refinement following use of the UCLA anisotropy server in two different cases. For one of them the Rfree went from 33% to 28%. I don't think it would have been publishable otherwise. Kendall

On May 1, 2012, at 11:10 AM, Bryan Lepore wrote:

...

On Mon, Apr 30, 2012 at 4:22 AM, Phil Evans wrote:

...

Are anisotropic cutoff desirable? is there a peer-reviewed publication - perhaps from Acta Crystallographica - which describes precisely why scaling or refinement programs are inadequate to ameliorate the problem of anisotropy, and argues why the method applied in Strong, et. al. 2006 satisfies this need?

-Bryan

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Hi Kendall, removing same amount of data randomly gives Rwork/Rfree ~ 30/35%. Pavel On 5/3/12 4:13 AM, Kendall Nettles wrote:

Hi Pavel, What happens if you throw out that many reflections that have signal? Can you take out a random set of the same size? Best, Kendall

On May 3, 2012, at 2:41 AM, "Pavel Afonine" wrote:

...

Hi Kendall,

I just did this quick test: calculated R-factors using original and anisotropy-corrected Mike Sawaya's data (*)

Original: r_work : 0.3026 r_free : 0.3591 number of reflections: 26944

Truncated: r_work : 0.2640 r_free : 0.3178 number of reflections: 18176

The difference in R-factors is not too surprising given how many reflections was removed (about 33%).

Pavel

(*) Note, the data available in PDB is anisotropy corrected. The original data set was kindly provided to me by the author.

On 5/2/12 5:25 AM, Kendall Nettles wrote:

...

I didnt think the structure was publishable with Rfree of 33% because I was expecting the reviewers to complain.

We have tested a number of data sets on the UCLA server and it usually doesn't make much difference. I wouldn't expect truncation alone to change Rfree by 5%, and it usually doesn't. The two times I have seen dramatic impacts on the maps ( and Rfree ), the highly anisotrophic sets showed strong waves of difference density as well, which was fixed by throwing out the noise. We have moved to using loose data cutoffs for most structures, but I do think anisotropic truncation can be helpful in rare cases.

Kendall

On May 1, 2012, at 3:07 PM, "Dale Tronrud" wrote:

...

While philosophically I see no difference between a spherical resolution cutoff and an elliptical one, a drop in the free R can't be the justification for the switch. A model cannot be made more "publishable" simply by discarding data.

We have a whole bunch of empirical guides for judging the quality of this and that in our field. We determine the resolution limit of a data set (and imposing a "limit" is another empirical choice made) based on Rmrg, or Rmes, or Rpim getting too big or I/sigI getting too small and there is no agreement on how "too big/small" is too "too big/small".

We then have other empirical guides for judging the quality of the models we produce (e.g. Rwork, Rfree, rmsds of various sorts). Most people seem to recognize that the these criteria need to be applied differently for different resolutions. A lower resolution model is allowed a higher Rfree, for example.

Isn't is also true that a model refined to data with a cutoff of I/sigI of 1 would be expected to have a free R higher than a model refined to data with a cutoff of 2? Surely we cannot say that the decrease in free R that results from changing the cutoff criteria from 1 to 2 reflects an improved model. It is the same model after all.

Sometimes this shifting application of empirical criteria enhances the adoption of new technology. Certainly the TLS parametrization of atomic motion has been widely accepted because it results in lower working and free Rs. I've seen it knock 3 to 5 percent off, and while that certainly means that the model fits the data better, I'm not sure that the quality of the hydrogen bond distances, van der Waals distances, or maps are any better. The latter details are what I really look for in a model.

On the other hand, there has been good evidence through the years that there is useful information in the data beyond an I/sigI of 2 or an Rmeas> 100% but getting people to use this data has been a hard slog. The reason for this reluctance is that the R values of the resulting models are higher. Of course they are higher! That does not mean the models are of poorer quality, only that data with lower signal/noise has been used that was discarded in the models you used to develop your "gut feeling" for the meaning of R.

When you change your criteria for selecting data you have to discard your old notions about the acceptable values of empirical quality measures. You either have to normalize your measure, as Phil Jeffrey recommends, by ensuring that you calculate your R's with the same reflections, or by making objective measures of map quality.

Dale Tronrud

P.S. It is entirely possible that refining a model to a very optimistic resolution cutoff and calculating the map to a lower resolution might be better than throwing out the data altogether.

On 5/1/2012 10:34 AM, Kendall Nettles wrote:

...

I have seen dramatic improvements in maps and behavior during refinement following use of the UCLA anisotropy server in two different cases. For one of them the Rfree went from 33% to 28%. I don't think it would have been publishable otherwise. Kendall

On May 1, 2012, at 11:10 AM, Bryan Lepore wrote:

...

On Mon, Apr 30, 2012 at 4:22 AM, Phil Evans wrote: > Are anisotropic cutoff desirable? is there a peer-reviewed publication - perhaps from Acta Crystallographica - which describes precisely why scaling or refinement programs are inadequate to ameliorate the problem of anisotropy, and argues why the method applied in Strong, et. al. 2006 satisfies this need?

-Bryan

---

phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

---

phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Hi Sebastiano, hm.. I don't know, I didn't have chance to think about it, I just found it attractive and easy to do Kendall's experiment, and post the results. Despite the obviousness of the result, something tells me that it's not that easy, but I'm not in the most convenient moment to carry out such a delicate process as thinking -;) I still don't like the idea of throwing data, even if some of list members believe it is not data but junk. That requires a thorough thinking and comprehensive testing: sounds like a project for someone to do. So, no - I would not take the result of this test as a green light to run all data sets through aniso-truncation server, no. What if you just start trying blindly removing reflections one by one and see if removing each one decreases the R, and actually remove one if and only if it does decrease R? That would be a way to come up with very robust and unjustifiable protocol of lowering R, but does anyone really want it? I think mathematicians should know methods to measure information content in the data, and that would probably a better route to take. Pavel

Hi Kendall, This could work. You could define a fixed set of test reflections, and never touch these, and never include them in refinement, and always use this fixed set to calculate a free R. Then you could do whatever you want, throw away some work reflections, etc, refine, and evaluate how things are working with the fixed free R set. All the best, Tom T ________________________________________ From: [email protected] [[email protected]] on behalf of Kendall Nettles [[email protected]] Sent: Thursday, May 03, 2012 7:05 AM To: PHENIX user mailing list Subject: Re: [phenixbb] Geometry Restraints - Anisotropic truncation Hi Pavel, Could you use a similar approach to figuring out where to cut your data in general? Could you compare the effects of throwing out reflections in different bins, based on I/sigma, for example, and use this to determine what is truly noise? I might predict that as you throw out "noise" reflections you will see a larger drop in Rfree than from throwing out "signal" reflections, which should converge as you approach the "true" resolution. While we don't use I/sigma exclusively, we do to tend towards cutting most of our data sets at the same i/sigma, around 1.5. It would be great if there was a more scientific approach. Best, Kendall On May 3, 2012, at 7:45 AM, Pavel Afonine wrote:

Hi Kendall,

removing same amount of data randomly gives Rwork/Rfree ~ 30/35%.

Pavel

On 5/3/12 4:13 AM, Kendall Nettles wrote:

...

Hi Pavel, What happens if you throw out that many reflections that have signal? Can you take out a random set of the same size? Best, Kendall

On May 3, 2012, at 2:41 AM, "Pavel Afonine" wrote:

...

Hi Kendall,

I just did this quick test: calculated R-factors using original and anisotropy-corrected Mike Sawaya's data (*)

Original: r_work : 0.3026 r_free : 0.3591 number of reflections: 26944

Truncated: r_work : 0.2640 r_free : 0.3178 number of reflections: 18176

The difference in R-factors is not too surprising given how many reflections was removed (about 33%).

Pavel

(*) Note, the data available in PDB is anisotropy corrected. The original data set was kindly provided to me by the author.

On 5/2/12 5:25 AM, Kendall Nettles wrote:

...

I didnt think the structure was publishable with Rfree of 33% because I was expecting the reviewers to complain.

We have tested a number of data sets on the UCLA server and it usually doesn't make much difference. I wouldn't expect truncation alone to change Rfree by 5%, and it usually doesn't. The two times I have seen dramatic impacts on the maps ( and Rfree ), the highly anisotrophic sets showed strong waves of difference density as well, which was fixed by throwing out the noise. We have moved to using loose data cutoffs for most structures, but I do think anisotropic truncation can be helpful in rare cases.

Kendall

On May 1, 2012, at 3:07 PM, "Dale Tronrud" wrote:

...

While philosophically I see no difference between a spherical resolution cutoff and an elliptical one, a drop in the free R can't be the justification for the switch. A model cannot be made more "publishable" simply by discarding data.

We have a whole bunch of empirical guides for judging the quality of this and that in our field. We determine the resolution limit of a data set (and imposing a "limit" is another empirical choice made) based on Rmrg, or Rmes, or Rpim getting too big or I/sigI getting too small and there is no agreement on how "too big/small" is too "too big/small".

We then have other empirical guides for judging the quality of the models we produce (e.g. Rwork, Rfree, rmsds of various sorts). Most people seem to recognize that the these criteria need to be applied differently for different resolutions. A lower resolution model is allowed a higher Rfree, for example.

Isn't is also true that a model refined to data with a cutoff of I/sigI of 1 would be expected to have a free R higher than a model refined to data with a cutoff of 2? Surely we cannot say that the decrease in free R that results from changing the cutoff criteria from 1 to 2 reflects an improved model. It is the same model after all.

Sometimes this shifting application of empirical criteria enhances the adoption of new technology. Certainly the TLS parametrization of atomic motion has been widely accepted because it results in lower working and free Rs. I've seen it knock 3 to 5 percent off, and while that certainly means that the model fits the data better, I'm not sure that the quality of the hydrogen bond distances, van der Waals distances, or maps are any better. The latter details are what I really look for in a model.

On the other hand, there has been good evidence through the years that there is useful information in the data beyond an I/sigI of 2 or an Rmeas> 100% but getting people to use this data has been a hard slog. The reason for this reluctance is that the R values of the resulting models are higher. Of course they are higher! That does not mean the models are of poorer quality, only that data with lower signal/noise has been used that was discarded in the models you used to develop your "gut feeling" for the meaning of R.

When you change your criteria for selecting data you have to discard your old notions about the acceptable values of empirical quality measures. You either have to normalize your measure, as Phil Jeffrey recommends, by ensuring that you calculate your R's with the same reflections, or by making objective measures of map quality.

Dale Tronrud

P.S. It is entirely possible that refining a model to a very optimistic resolution cutoff and calculating the map to a lower resolution might be better than throwing out the data altogether.

On 5/1/2012 10:34 AM, Kendall Nettles wrote:

...

I have seen dramatic improvements in maps and behavior during refinement following use of the UCLA anisotropy server in two different cases. For one of them the Rfree went from 33% to 28%. I don't think it would have been publishable otherwise. Kendall

On May 1, 2012, at 11:10 AM, Bryan Lepore wrote:

> On Mon, Apr 30, 2012 at 4:22 AM, Phil Evans wrote: >> Are anisotropic cutoff desirable? > is there a peer-reviewed publication - perhaps from Acta > Crystallographica - which describes precisely why scaling or > refinement programs are inadequate to ameliorate the problem of > anisotropy, and argues why the method applied in Strong, et. al. 2006 > satisfies this need? > > -Bryan

---

phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

---

phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

1) maps calculated using all (unmodified) data by phenix.refine, phenix.maps and similar tools are better than maps calculated using anisotropy truncated data. So, yes, for the purpose of map calculation there is no need to do anything: Phenix map calculation tools deal with anisotropy very well.

That is not our experience in cases of really severe anisotropy.

2) phenix.refine refinement may fail if one uses original anisotropic data set. This is probably because the ML target does not use experimental sigmas (and anisotropy correction by UCLA server is nothing but Miller index dependent removing the data by sigma criterion - yeah, that old well criticized practice of throwing away the data that you worked hard to measure!). May be using sigmas in ML calculation could solve the problem but that has to be proved.

UCLA server removes all the data beyond set ellipsoid, it does not deal with individual reflections by sigma. Also one can set its own resolution limits on UCLA server, depending on personally preferred criteria. As for throwing away data, what is the difference here from "throwing away data" when we cut the resolution isotropically (one high res limit) using whatever criterion (I/sigma, R, correlation) one prefers? Cheers Leonid

...

...

1) maps calculated using all (unmodified) data by phenix.refine, phenix.maps and similar tools are better than maps calculated using anisotropy truncated data. So, yes, for the purpose of map calculation there is no need to do anything: Phenix map calculation tools deal with anisotropy very well.

That is not our experience in cases of really severe anisotropy.

Can you send me an example off-list, please (need model and data files (before and after truncation)).

The improvement after truncation is consistent over many different datasets, but the effects are sometimes subtle, so I will have to look for a clear-cut example when I have time. Truncation also helps a lot for density modification. Don't think I ever seen maps getting worse after truncation (although we usually use a bit more generous limits than UCLA server suggests).

...

...

2) phenix.refine refinement may fail if one uses original anisotropic data set. This is probably because the ML target does not use experimental sigmas (and anisotropy correction by UCLA server is nothing but Miller index dependent removing the data by sigma criterion - yeah, that old well criticized practice of throwing away the data that you worked hard to measure!). May be using sigmas in ML calculation could solve the problem but that has to be proved.

UCLA server removes all the data beyond set ellipsoid, it does not deal with individual reflections by sigma.

Yes, it's not done per reflection, but indirectly as part of determination of the parameters of that ellipsoid that is then used to cut the data (as far as I understand it).

...

Also one can set its own resolution limits on UCLA server, depending on personally preferred criteria.

May be it's just fine (given the current state-of-the-art of methods used in refinement) if it's done carefully and thoughtfully. The trend though seem to be to blindly use it "just in case it gives me a lower R", which I find dangerous (and yes, it is wrong to compare R-factors calculated using different amount of data!).

Pavel

Of course the main criterion for using truncation is improved (or not) appearance of maps and ability to continue building/refining into those maps. R factors do not drop that much due to truncation per se. Leonid