Dear Gerard,
A simple compromise could be to report your statistics to the resolution limit in your worst direction (so completeness and other statistics will not have started to really suffer from excluded reflections) and then in your best direction (where completeness will be abysmal), or what is the same, for the complete dataset. The resolution limits implied would be given in the legend or in a similar 2-number format in the resolution line of table1.
There is a lot of information missing from a complete tensor, but most of what most people would ever need to know (what are the effective resolution and quality of the data?) is already captured there. The big advantage is that people are already used to reporting two numbers in each cell of table 1, one for the full dataset and one for the high resolution shell. This would be no different, it would just use different resolution cutoffs.
Just my 2 cents.
Best,
Jose.
================================
Jose Antonio Cuesta-Seijo, Ph.D.
Carlsberg Research Laboratory
Gamle Carlsberg Vej 10
DK-1799 Copenhagen V
Denmark
Tlf +45 3327 5332
Email [log in to unmask]
================================
-----Original Message-----
From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of Gerard Bricogne
Sent: Thursday, February 18, 2016 8:30 PM
To: [log in to unmask]
Subject: Re: [ccp4bb] Reporting resolution and completeness for anisotropic data
Dear all,
I had a very interesting reply (off-list) from Oliver Clarke at Columbia University who directed me to a recent paper
http://science.sciencemag.org/content/351/6273/608.full-text.pdf+html
in which Table I was extended with extra rows to give some details about the resolution limits in different directions according to two criteria. It is good that this was accepted by the reviewers and the journal as useful information. This extension, however, was designed by the authors rather than formatted according to a standard. It is also a relatively simple case, as the crystal is orthorhomic, so that the principal directions are the cell axes themselves and do not have to be specified by strange fractional components along those axes as would be the case with a monoclinic or triclinic cell.
I was not referring to standards out of a love for committees and guidelines, but if this information is presented in different ways by different authors in different journals, it will be very difficult to harvest it and include it in the corresponding PDB entries. In the case of this paper, the PDB files for the two structures that have anisotropic data contain no trace of the careful description of that anisotropy given in the paper.
Beyond the question of Table I (which seems to be more elastic than I was aware of) lies the more far-reaching one of coming up with appropriate descriptors of anisotropy and completeness and making them into standard items of information in results of data processing and in PDB depositions.
With best wishes,
Gerard.
--
On Wed, Feb 17, 2016 at 06:21:42PM +0000, Gerard Bricogne wrote:
> Dear Colleagues,
>
> Since the announcement of the STARANISO server in the message
> included below, numerous groups have submitted datasets for anisotropy
> characterisation and optional correction, and sent us feedback. A few
> cases are illustrated on the "Gallery" at
>
> http://staraniso.globalphasing.org/staraniso_gallery.html
>
> and more will be posted.
>
> One of these users, who saw substantial improvements in his maps,
> has asked us the natural question: what should I put in Table I of my
> manuscript concerning the resolution and the completeness of my data
> after treatment with STARANISO?
>
> As usual, anisotropy needs several numbers to describe it rather
> than just one, and the problem is that Table I has only one cell for
> resolution. Besides requiring up to three distinct numbers for the
> resolution limits in several "principal" directions, it is necessary
> to specify these directions themselves. At first sight one might think
> that these will be the reciprocal cell axes, but as shown by the third
> example in the Gallery, this may not be the case in low-symmetry space
> groups.
>
> There is "prior art" in this area, for instance in the AIMLESS
> program that reports resolution limits in general reciprocal-space
> directions if necessary, and we have checked that these agree with
> those produced by STARANISO. The main point is: how can one fit all
> this information into Table I without revisiting its format? And, by a
> knock-on effect, whatever goes into Table I should end up in the PDB
> entry for the published structure: how, then, should room be made so
> that resolution can be upgraded from a scalar to at least a 2nd-order
> tensor?
>
> A related question is that of coming up with a meaningful
> definition of completeness in the presence of anisotropy. A certain
> notion of incompleteness still needs to be associated with that of
> missing regions (e.g. angular wedges) of reciprocal space, simply
> because they were not collected, and therefore to the idea of being
> able to increase it through a better executed experiment (e.g taking
> the trouble to "fill the cusp"). When anisotropy is present, one is
> forced by the current conventions to declare completeness of the set
> of unique reflections for which significant intensities are available
> (rather than just "indices", to quote a famous witticism by Keith
> Wilson) by reference to the resolution limit in the best direction(s).
> This can result in alarmingly low values, even though everything worth
> measuring has been measured.
>
> Shouldn't there be a way in which a dataset containing all
> significant data measurable on a crystal with anisotropic diffraction
> limits would be credited as having 100% completeness, as there is no
> way of measuring more? With the present definitions it is impossible
> to tell whether one is missing indices for which intensities would
> have no information content anyway (because of anisotropy) or indices
> for which significant intensities would have been expected but didn't
> get measured (because of a sub-optimal strategy or by lack of care).
>
> At least two different numbers would therefore seem to be needed:
>
> (1) completeness with respect to the resolution limit in the best
> direction(s), so that a "data miner" doesn't get misled as to the
> accuracy of the coordinates;
>
> (2) completeness with respect to e.g. an ellipsoid-like region to
> which anisotropy restricts measurable reflections, so that credit is
> given to the experimentalist for having done his/her best, given the
> circumstances.
>
>
> Many of you will have thoughts on this: we look forward to
> hearing them!
>
>
> With best wishes,
>
> Gerard, Ian & the Global Phasing developers.
>
> --
> On Thu, Jan 07, 2016 at 10:56:04AM +0000, Gerard Bricogne wrote:
> > Dear Colleagues,
> >
> > We have been working for quite some time on new software
> > (called
> > STARANISO) to analyse the anisotropy of X-ray diffraction intensity
> > data and investigate the appropriateness of various remedial
> > measures, as well as to revisit subtleties in standard processing
> > steps such as TRUNCATE where the existing treatment left something
> > to be desired in the presence of significant anisotropy.
> >
> > We are pleased to invite you to test-drive this program via a
> > Web server that offers access to it, and would greatly value your
> > feedback on it. To do this, please connect to
> >
> > http://staraniso.globalphasing.org
> >
> > and peruse the material, links and buttons presented to you there.
> >
> > As it is the whole capability (including its user
> > documentation) that is being submitted to your scrutiny, we will say
> > no more, as any obscurity experienced by a first-time would-be user
> > is part of what we are counting on you to report to us :-) .
> >
> > Thank you in advance for your time and attention in evaluating
> > this server and the usefulness of what STARANISO produces for you.
> > We very much look forward to your feedback and suggestions.
> >
> >
> > While this request can be found at the bottom of the main
> > server Web page, we expressly ask here again that you send your
> > feedback to
> >
> > [log in to unmask]
> >
> > and not to individual developers that you might know. Thank you!
> >
> >
> >
> > Wishing you a Happy and more isotropic New Year,
> >
> > Ian Tickle & the Global Phasing developers.
