Biomedical Engineering Reference
In-Depth Information
the control samples also exhibit the same effect, showing that the variation
is almost certainly due to spectrometer drift. The data have not yet been
normalised to the internal standards which were spiked into the samples,
an operation for the moment that has to be done outside mzMine. By
removal of the control samples some separation of wild type from mutant
was observed (not shown). Sammon's projection (Figure 4.12(d)) is a non-
linear multidimensional scaling method which projects multidimensional
data down to just two dimensions. It is a useful as a method to examine
approximate clustering in data but offers no useful interpretability.
4.5.2 Other features in mzMine
There are many more features in mzMine, including some support for
ms/ms data and extensive searching of external internet databases. More
features are being added all the time, including several experimental baseline
correction algorithms. Development in this area includes a raw data baseline
correction module based on asymmetric least squares from the R package
ptw: parametric time-warping [29] and an interface to the NIST MS Search
[30] program to allow the use of mzMine for GC-MS data.
4.6 Metabolomics data processing using
the open source workfl ow engine, KNIME
Data output from metabolomics software almost always requires further
formatting before the data analysis stage. Procedures such as normalising
the data to internal standards, or to total signal and subtotalling adducts
are possible using spreadsheets and manual manipulation. The
disadvantage of this approach is the ability to make unintentional errors
and also the lack of transparency as to what was done to the data.
An alternative to this approach is to use a workfl ow tool which uses
small data processing nodes linked together and therefore inherently
documents the data set operations such as the open source project
KNIME [31, 32]. Chapter 6 by Meinl, Jagla and Berthold describes
applications of KNIME in chemical and bioinformatics. KNIME was
developed by the Centre for Bioinformatics and Information Mining at
the University of Konstanz and is based on the Eclipse platform for Java
[33]. The KNIME software comes with a number of standard nodes but
also there is a growing community of both commercial and open source
developers writing new nodes for many data processing tasks. Included
Search WWH ::
Custom Search