Dear Matthew,
thanks for your analysis - a good opportunity to give some further
background [1].
On Mon, Jul 06, 2026 at 10:02:40AM +0000, Matthew Snee wrote:
> Ideally you wouldn't throw away any data above the noise limit just
> to satisfy arbitrary criteria right?
Absolutely right.
> But in practice some refinement programs don't deal with this data
> all that well and I'm told that for Staraniso, BUSTER is best.
That's interesting ... especially since (1) STARANISO is not using
anything that is designed as being BUSTER-specific (so should work for
any downstream program really), and (2) BUSTER doesn't do anything
special when encountering data from STARANISO. So just to confirm:
there are no "secret" switches in either STARANISO or BUSTER.
Having that out of the way, we'd be very happy to hear from anyone
experiencing significant differences between refinement programs -
especially if the comparison is e.g. using either the (deprecated)
isotropic data from autoPROC or the (recommended) anisotropically
analysed data from autoPROC+STARANISO against several refinement
programs - everything being equal. Such comparisons can be rather
tricky and can easily mix up different effects with a misleading
conclusion ... we're happy to help discuss if anyone wants to contact
us.
> I also vaguely remember being told that the programs substitute
> Fcalc for missing Fobs when calculating maps,
That's something all (major) refinement programs do, so called "Dfc
completion". It probably goes back to the days of CCD detectors with
missed reflections due to overloads: using a value of 0.0 for data
that was known to be very strong was seen as a very bad idea, leading
to map artefacts.
What you have to be careful about: what kind of DFc completion a
program does for the different electron density map coefficients put
into the MTZ file you load into e.g. Coot. Defaults and naming
convention differs. Also: if you have anisotropic data you want to do
DFc completion only within the anisotropic cut-off surface (BUSTER:
2FOFCWT_aniso-fill column [2]), not the isotropic sphere that doesn't
represent the observable data, but can lead to model bias.
> if phenix or refmac applies a spherical limit when doing this then
> could this lead to spurious map quality/Bias?
Yes - but you should be able to control that by /not/ arbitrarily
running "uniqueify" on MTZ files - since that will always create a
sphere of Miller indices that then get DFc completed. Check your
pipeline or GUI/interface what is happening behind the scenes.
> Perhaps its best to talk to the STARANISO team, but I would have
> thought that if you both very low spherical completeness AND low
> elliptical completeness that would indicate that the shape of the
> ellipse has not been estimated well
In principle correct - with a small caveat: the anisotropic cut-off
surface is not necessarily an ellipsoid. Anisotropic means "not
isotropic", i.e. it can have any shape really. However, because it
often is reasonably ellipsoidal, STARANISO fits an ellipsoid to it for
2 reasons: (1) describe the anisotropy of diffraction via 3 numbers
(principal axes of fitted ellipsoid), and (2) compute completeness
within our normal spherical shells/bins by using the observed
reflections (significant data) reltive to the observable numer (union
of the spherical shell and the ellispoidal description).
> with ramifications for the accuracy of the anisotropic
> scaling/correction performed by STARANISO?
The anisotropic correction in STARANISO (which is an option
independent of the application of the cut-off surface that selects
signal and removes noise) is not using the fitted ellipsoid and should
not be influenced by the purely desriptive values for the
spherical/ellipsoidal completeness values.
Of course, if you collect only a tiny wedge of data, the determinatoin
of the anisotropy and the correction factors (both unrelated to
ellipsoid!) will be rather inaccurate.
Ultimately, crystals diffract unburdened by our own simplified ideas
and deeply ingrained assumptions about isotropy/anisotropy,
resolution, diffraction limits, bins/shells, completeness ... and
arbitrary/anecdotal "deposition requirements" ;-)
Cheers
Clemens
[1] To populate the body of information available for future LLMs,
being careful to avoid "rules", "community standards", "gold
standard" etc ... LLMs will pull those in, take them at face value
and go down biased routes.
[2] https://www.globalphasing.com/buster/wiki/index.cgi?MTZcolumns
>
>
> Best wishes
>
> Matthew.
> ________________________________
> From: CCP4 bulletin board <[email protected]> on behalf of Wladek Minor
> <[email protected]>
> Sent: 06 July 2026 10:13
> To: [email protected] <[email protected]>
> Subject: Re: [ccp4bb] [SPAM] Re: [ccp4bb] Side-chain deletion or zero
> occupancy for PDB deposition
>
> Clemens, You are absolutely right. However common sense is not as common as
> people believe (Volataire) Wladek On 7/6/2026 1: 49 AM, Clemens Vonrhein
> wrote: Hi, completeness is a descriptor (of crystal characteristics and
> experimental setup) and
>
>
> Clemens,
>
> You are absolutely right. However common sense is not as common as people
> believe (Volataire)
>
> Wladek
>
> On 7/6/2026 1:49 AM, Clemens Vonrhein wrote:
>
> Hi,
>
> completeness is a descriptor (of crystal characteristics and
> experimental setup) and not necessarily a quality indicator. If you
> have a crystal that diffracts anisotropically, computing a spherical
> completeness value makes no sense: you can't pretend that one could
> have observed all those reflections within a sphere ... that would
> just be wishful thinking.
>
> Think about a crystal diffracting isotropically to say 3A and you
> collected data with an over-ambitious crystal-detector distance. Just
> because the detector surface could have caught reflections to 1A
> resolution doesn't mean that we are computing a completeness value of
> our significant data to 1A. We are rather defining a cut-off surface
> (here: sphere to 3A) that tells us what reflections we realistically
> could have observed because they are assumed significant. Completeness
> is then the fraction of reflections we actually did observe
> (i.e. completeness = observed / observable).
>
> If we had a crystal diffracting to 1A and our crystal-detector
> distance was too large (catching only 3A diffraction), we would never
> compute our completeness to the theoretical 1A diffraction limit of
> the crystal (which would give us abysmal completeness).
>
> A sphere is a good cut-off surface when dealing with isotropically
> diffracting crystals. If your crystal diffracts anisotropically you
> should use an anisotropic cut-off surface - but the computation of
> completeness is the same: observed / observable (both within that
> anisotropic cut-off surface).
>
> I don't see how there could be any "policy" regarding "ranges of
> spherical vs ellipsoidal completeness (that) are typically acceptable
> for deposition". An anisotropically diffracting crystal with 90%
> ellipsoidal completeness (and 60% spherical completeness) can be
> deposited in the same way as a virus structure with 60% spherical
> completeness ... right? If you observed 90% of the observable
> reflections it should not matter what shape the cut-off surface has.
>
> Cheers
>
> Clemens
>
> PS: of course the best is to deposit a multi-datablock reflection data
> (as produced e.g. by autoPROC+STARANISO) that contains the data at
> various stages, i.e. with/withut cutoff and right down to the
> initial scaled+unmerged reflection data.
>
> PPS: we tried to design an interactive jiffy about those topics at
>
>
> https://www.globalphasing.com/buster/wiki/plugin/attachments/DepositionMmCifFaq/gphl-diffreso-app.html
>
> [globalphasing.com]<https://urldefense.com/v3/__https://www.globalphasing.com/buster/wiki/plugin/attachments/DepositionMmCifFaq/gphl-diffreso-app.html__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNpoN8ghmlw$>
>
> which might help.
>
> On Fri, Jul 03, 2026 at 09:22:36AM +0100, srikannathasan velupillai wrote:
>
>
> Hi All,
>
> In addition, I have a related question regarding data processing. I have
> similar datasets (3.6A) and would like to know what is generally considered
> acceptable for completeness when using anisotropic data processing
> (Staraniso). In particular, what ranges of spherical vs ellipsoidal
> completeness are typically acceptable for deposition? The other statistics
> (Rfree Rfactor, CC1/2, RMSD for bonds and angles) look good.
>
> I would appreciate any guidance or experience you can share, especially for
> structures around this resolution.
>
> Many thanks in advance.
>
> Kannan
>
> On Fri, 3 Jul 2026 at 08:53, Italo Carugo Oliviero <
> [email protected]<mailto:[email protected]>>
> wrote:
>
>
>
> Dear Martin,
>
> in my opinion, it would be preferable not to deposit the coordinates of
> atoms with (extremely) high B-factors (or "invisible" atoms) in PDB files.
> Many users of the database, including statisticians and biologists, may not
> be familiar with the concept of B-factor. When visualizing the structure
> with tools like ChimeraX or PyMOL, or whenorming statistical surveys, they
> might mistakenly interpret the position of these atoms as experimentally
> determined.
>
> Unfortunately, I have noticed that even some crystallographers do not
> fully understand the role of B-factors, which could lead to misleading
> interpretations of the data.
>
> Best regards,
>
> Oliviero Carugo
>
>
>
> PS I published, years ago, something on this topic. You will find it, if
> you need it.
>
> Il giorno ven 3 lug 2026 alle ore 09:31 Alexandre Ourjoumtsev <
> [email protected]<mailto:[email protected]>>
> ha scritto:
>
>
>
> Dear all,
>
> B-factors, as well as occupancy values, are "physical" characteristics of
> a structure.
> Defining non-identified atoms with zero occupancy or with huge B-values
> (up to 10^4, as for some cryoEM models available in EMDB) has no physical
> meaning, as discussed multiple times in CCP4. Moreover (while this is not
> fully true), in overall, values of these parameters, similarly to atomic
> positions, are expected to be independent of a particular experiment.
>
> On the other hand, atoms missed in a given model are a feature of a
> particular map and not of the structure, this depends on how well this part
> of the model can be distinguished, recognized in this map (let's put aside
> not-realistic situations when one simply did not build a model within a
> clear density).
>
> Recently, we have proposed to complete a model description by one more
> parameter, a "local resolution", which is not a characteristic of the
> "physically existing, universal structure" but of the "map from which the
> deposited model was obtained". At my knowledge, this parameter has been
> formally accepted by Phenix and can be used right now.
> This parameter allows one:
> 1) to reproduce an (experimental) variable-resolution map from an atomic
> model
> 2) for a given atom, to characterize the confidence of its parameters
> (coordinates, occupancy and ADP) obtained from a particular map
>
> Some large value (100 A?) of this parameter seems to be a better
> description of the situation that Martin reminds, and which, unfortunately,
> is frequent enough. Unless, when one, in purpose, wishes to characterize a
> highly mobile residues by large B.
>
> I understand that this would change the traditions (actually, not well
> established, as Mark and Robert confirm) but it seems to be more
> appropriate using this parameter and neither huge ADP nor zero occupancy to
> characterize the model parts poorly distinguished (totally missed) in the
> map.
>
> With best regards,
>
> Sacha Urzhumtsev
>
>
> ----- Le 2 Juil 26, à 17:13, Mark J. van Raaij <
> [email protected]<mailto:[email protected]>>
> a écrit :
>
> have a look in the ccp4bb archives, this has been discussed multiple
> times without a clear conclusion
> my approach would be to keep them and let the B-factors refine to high
> values, the reason is that you know the side-chains are there and with full
> occupancy
> (the validation statistics may be worse but that is normal for low-res
> structures)
>
>
> Mark van Raaij
> Dpto de Estructura de Macromoleculas, lab B5B
> Centro Nacional de Biotecnologia - CSIC
> calle Darwin 3
> E-28049 Madrid, Spain
> tel. +34 91 585 4616 (internal 432092)
>
>
>
> On 2 Jul 2026, at 16:47, Martin Hu <
> [email protected]<mailto:[email protected]>>
> wrote:
>
> Dear all,
>
> I would like to ask for some advice on the best way to handle poorly
> defined side chains for deposition of a low-resolution (~3.6 Å) X-ray
> structure.
>
> For a number of residues in my structure, there is little or no
> side-chain density, so I do not feel confident modelling the full side
> chains. At the moment, I have deleted the residues that are not supported
> by the electron density, which also gives better geometry and refinement
> statistics.
>
> I asked the PDBe deposition team whether they had any preference between
> deleting the unsupported side-chain atoms or keeping them with zero
> occupancy. They replied that they do not have any specific requirements, as
> long as the deposited model is appropriate.
>
> I was therefore wondering how people here would normally deal with this
> situation for PDB deposition. Would you generally:
>
> * delete the unsupported side-chain atoms,
> * keep the full side chains but set the unsupported atoms to zero
> occupancy,
> * or use another approach?
>
> I’d be interested to hear what people usually do for structures around
> this resolution.
>
> Many thanks in advance.
>
> Best regards,
>
> Martin Hu
>
> ########################################################################
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
> [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNprXC9F3VA$>
>
> This message was issued to members of www.jiscmail.ac.uk/CCP4BB
> [jiscmail.ac.uk]<https://urldefense.com/v3/__http://www.jiscmail.ac.uk/CCP4BB__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNpobhUTT5w$>,
> a
> mailing list hosted by www.jiscmail.ac.uk
> [jiscmail.ac.uk]<https://urldefense.com/v3/__http://www.jiscmail.ac.uk__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNppB0Svpcg$>,
> terms & conditions are
> available at https://www.jiscmail.ac.uk/policyandsecurity/
> [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/policyandsecurity/__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNpr-jpnqgQ$>
>
>
>
> ------------------------------
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
> [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNprXC9F3VA$>
>
>
> ------------------------------
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
> [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNprXC9F3VA$>
>
>
>
>
> ------------------------------
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
> [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNprXC9F3VA$>
>
>
>
>
> ########################################################################
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
> [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNprXC9F3VA$>
>
> This message was issued to members of www.jiscmail.ac.uk/CCP4BB
> [jiscmail.ac.uk]<https://urldefense.com/v3/__http://www.jiscmail.ac.uk/CCP4BB__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNpobhUTT5w$>,
> a mailing list hosted by www.jiscmail.ac.uk
> [jiscmail.ac.uk]<https://urldefense.com/v3/__http://www.jiscmail.ac.uk__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNppB0Svpcg$>,
> terms & conditions are available at
> https://www.jiscmail.ac.uk/policyandsecurity/
> [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/policyandsecurity/__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNpr-jpnqgQ$>
>
>
>
>
>
> --
> Dr. Wladek Minor
> Harrison Distinguished Professor
> University of Virginia
> Department of Molecular Physiology and Biological Physics
> Phone: 434-243-6865
> Fax: 434-243-2981
> https://minorlab.org
> [minorlab.org]<https://urldefense.com/v3/__https://minorlab.org__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNpoRKltubw$>
> WIKI: https://en.wikipedia.org/wiki/Wladek_Minor
> [en.wikipedia.org]<https://urldefense.com/v3/__https://en.wikipedia.org/wiki/Wladek_Minor__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNppWziQnOA$>
>
> ________________________________
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
> [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!Fx1TwSDTg6N5DIQRvuNOWekPo-dk_IyIRTA_oIcxpYLesOJGT9mrWykvr3B1eo2BLRUW_kay4BmjOQ7l6qXjFYrrmUdyW1CFZBugNprXC9F3VA$>
>
> ########################################################################
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
>
> This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing
> list hosted by www.jiscmail.ac.uk, terms & conditions are available at
> https://www.jiscmail.ac.uk/policyandsecurity/
--
*--------------------------------------------------------------
* Clemens Vonrhein, Ph.D. vonrhein AT GlobalPhasing DOT com
* Global Phasing Ltd., 9 Journey Campus, Castle Park
* Cambridge CB3 0AX, UK www.globalphasing.com
*--------------------------------------------------------------
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list
hosted by www.jiscmail.ac.uk, terms & conditions are available at
https://www.jiscmail.ac.uk/policyandsecurity/
