Cameron Neylon

During the [workshop in late
February](http://blog.openwetware.org/scienceintheopen/2008/02/29/placeholder-post-for-workshop-comments/)
we had discussions about possible implementations of
[Taverna](http://taverna.sourceforge.net/) work flows to automate
specific processes to make our life easier. One specific example we
discussed was the reduction and initial analysis of Small Angle Neutrons
Scattering data. Here I want to describe a bit of the background to what
this is and what we might do to kick of the discussion.

Small angle scattering (SAS) is a technique that can be used to
determine large scale structural features (1 -- 100s nm) of molecules in
solution. This makes it appealing for the study of complexes, partially
and completely disordered systems, as well as confirming that the
overall solution structure is similar to that determined from high
resolution techniques. For biomolecular structure it is essentially
possible to get shapes at some level of detail. Most SAS is carried out
using X-rays but there are specific advantages to using neutrons for
some systems, particularly biological complexes (I've written a recent
[review on SAXS/SANS](http://www.ncbi.nlm.nih.gov/pubmed/18214466) in
biology if you want more details).

SAS does what is says on the can. You put a sample in a collimated beam
and observe the scattering of xrays or neutrons at small angles, usually
less than 5 deg. The general setup involve a sample position in front of
a long tank with a planar detector that can be moved up and down the
tank. For SANS there are a range of background and normalisation
measurements that need to be carried out to characterise absorption and
other issues. This means that converting the area detector data
(basically an image file, usually of around 1k x 1k pixels) to a simple
plot of scattered intensity (I) vs angle (usually given in units of
momentum transfer, Q\[angstroms\^-1\]) is not trivial. However it can
usually be reduced to a set of measurements that need to be done to
characterise the state of the instrument (which generally doesn't change
for a given set of measurments) and the measurements on the sample
itself. In addition to this there are (nearly) as many different tools
for converting area detector data to I vs Q as there are people looking
after these instruments.

What we would like to do is to be able to load up all of our raw data
(or at least pointers to it) into the LaBLog and then execute a
[Taverna](http://taverna.sourceforge.net/) workflow (which we could
share in [MyExperiment](http://www.myexperiment.org/)) to process, or
help us process, the data in a semi-automated fashion. This is assisted
by the fact that one of the data reduction tools,
[GRASP](http://www.ill.eu/sites/grasp/grasp_main.html) developed by
Charles Dewhurst at the [ILL](http://www.ill.fr), is written in MatLab.
Since the data we have to hand is ILL data from the
[D22](http://www.ill.eu/d22/home-d22/) instrument I will focus on this.

The idea would be, for a specific set of data, to define those data
files that relate to the instrument configuration and those that relate
to samples. [GRASP](http://www.ill.eu/sites/grasp/grasp_main.html)
allows a set of data files to be loaded in that define the
configuration. At this point there is manual intevention required to
define beam centres and to identify any dodgy pixels or detector
elements. There are two main types of files, transmission data involves
putting an attenuator in the beam and letting the direct beam hit the
detector, scattering data is taken with the attenuator out, but a beam
stop in to protect the detector from the direct beam.

The data files required for characterising a given instrument
configuration are; transmission and scattering from the empty beam (no
sample), transmission and scattering from an empty sample cell,
transmission and scattering from an H2O sample (sometimes cadmium gives
a 'flat background' for normalising between experiments), and
transmission and scattering from a boronated carbon sample (B4C,
sometimes vanadium, essentially blocks the beam and gives the
'electronic background', i.e. the detector background). Once this is set
up the processing of data files for that specific instrument
configuration is relatively straightforward

For each sample scattering and transmission data are recorded. These are
then used, with the instrument configuration setting to generate a
straightforward set of one dimensional data (I vs Q, error in I) which
is the data which is generally used for further analysis. This reduction
of the sample data can be fairly readily automated. We also have a
recently agreed [XML format for 1-D reduced small angle scattering
data](http://www.smallangles.net/wgwiki/index.php/cansas1d_documentation),
so we have a nice format to output.

The idea is therefore as follows:

1.  A [Taverna](http://taverna.sourceforge.net/) workflow is invoked
    that requires pointers to the instrument configuration files
    required. These could be found either by user instruction, or via
    the appropriate pointers in the LaBLog
2.  The workflow obtains the files and calls the MatLab code, loading up
    the files into the correct slots.
3.  It then calls the user to check over details that need manual
    intervention, such as beam centres, and any dodgy pixels. This is
    done via the GUI of the MatLab code in
    [GRASP](http://www.ill.eu/sites/grasp/grasp_main.html). The
    instrument configuration setup is then finished.
4.  The workflow then grabs the appropriate data files and automatically
    processes them, generating the appropriate reduced data.
5.  [Taverna](http://taverna.sourceforge.net/) then writes to the LaBLog
    a record of what it has done including pointers to input and output
    files, processing parameters etc.

This potentially automates a lot of the grunt work without removing the
checks that are required so as not to generated badly reduced data. In
fact if the workflow asks questions along the way it can asist the user
with doing the processing correctly. Finally the workflow can be readily
shared using [MyExperiment](http://www.myexperiment.org/), although
there is still likely to be the requirement to download
[GRASP](http://www.ill.eu/sites/grasp/grasp_main.html).


During the workshop in late February we had discussions about possible implementations of Taverna work flows to automate specific processes to make our life easier. One specific example we discussed was the reduction and initial analysis of Small Angle Neutrons Scattering data. Here I want to describe a bit of the background to what this is and what we might do to kick of the discussion.

Messaggi di Rogue Scholar

Incorporating My Experiment and Taverna into the LaBLog – A possible example