[buster-discuss] two questions : one about negative reflections and one about Simulated annealing

[Ian Tickle]

Hi Nicolas

The Bayesian method is in common use because it is the only method that
does not result in biased Fs (and therefore biased electron density), as
discussed in the article I linked to.

Cheers

-- Ian


On Tue, 22 Jun 2021 at 11:36, Ian Tickle <ianjt05 at gmail.com> wrote:

>
> Hi Nicolas
>
> As far as I know, Buster doesn't have a Simulated Annealing option, but
> maybe the Buster developers will have more to say on that.
>
> On your second question, there's an important step between getting the
> merged intensities and performing the refinement with Buster.  Again as far
> as I know Buster only uses amplitudes (Fs) as input.  Fs unlike
> intensiities cannot be non-negative by definition.  So we need to convert
> all the merged intensities including negative ones to positive Fs.  This is
> usually done by French & Wilson Bayesian estimation:
>
>
> https://warwick.ac.uk/fac/sci/statistics/staff/academic-research/french/publications/bayes_impact_french_wilson.pdf
>
> Also see the reference to the original 1978 paper cited there.
>
> This algorithm, or a modification of it, is implemented, among others, in
> the Truncate, CTruncate and StarAniso programs.  For example, StarAniso
> computes the Fs using a very accurate numerical method instead of using
> table look-ups, and takes into account the effect of anisotropy on the
> Bayesian prior.
>
> Best wishes
>
> -- Ian
>
>
> On Tue, 22 Jun 2021 at 10:29, Nicolas Foos <nicolas.foos at ibs.fr> wrote:
>
>> Dear buster users and developers,
>>
>> I have two questions : one simple : is it possible to perform a simulated
>> annealing using buster. I didn't see anything like that in the manual, but
>> maybe I miss it.
>>
>> The second question : how autoBuster handle the negative reflections
>> present in the data file ?
>> - Does it ignore them ?
>> - Does it fill them ?
>> - Does it set them to zero ?
>> What's you opinion about that point, should we use them or not. I would
>> be glad to have your opinion about that point.
>>
>> To let you know a bit more : I am working on time resolve project. The
>> Data were originally collected in XFEL, then processed using CrystFel. And
>> then based on Fo(state1)-Fo(state2) difference map calculation, I did an
>> extrapolation to refine the model. The data which generates my questions
>> are the Extrapolated ones.
>>
>> I am fully open to any comments.
>>
>> All the best,
>>
>> Nicolas Foos
>>
>> _______________________________________________
>> buster-discuss mailing list
>> buster-discuss at globalphasing.com
>> https://www.globalphasing.com/mailman/listinfo/buster-discuss
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.globalphasing.com/pipermail/buster-discuss/attachments/20210622/0b1feaa4/attachment-0001.html>