FiatFlux-P - User-Level Workflow Design: A Bioinformatics Perspective - page 107

5.2.1 FiatFlux Services

Turning FiatFlux' components into Bio-jETI-compliant services required
some major changes to its Matlab code base, in particular the rigorous re-
moval of all code for graphical components and the replacement of invocations
from the GUI by (parameterized) functions that can be called from an exter-
nal application. The user interactions required during the FiatFlux analysis
process can be categorized into simple input of experimental data or the se-
lection of modeling parameters and more intricate user interactions targeted
to the optimization of the calculations. Whereas the data input could easily
be replaced by parameterized functions, for the other steps the expert logic
had to be translated into quantifiable criteria. The functions developed to em-
ulate these tasks are packaged into two new modules ratio guiemulation
and netto guiemulation , respectively (detailed descriptions of the emula-
tion functions and some other additionally implemented functionality can be
found in [89, Chapter 6].
Next, the newly implemented functions were combined with the standard
FiatFlux functions in order to provide self-contained functional units. These
were then integrated and turned into SIBs using the jETI technology (cf.
Section 2.1.2). The following SIBs are central in FiatFlux-P:

•

ConvertMS converts mass spectrometry data from netCDF format to the
FiatFlux format.

• Ratio performs a predefined, complete flux ratio computation.

•

Netto performs a predefined, complete net flux distribution computation.

•

Netto CustomModel performs a predefined, complete net flux distribution
computation based on a user-defined network model.

•

Netto JointRatios performs a predefined, complete net flux distribution
computation based on the combined flux ratios from two different exper-
iments.

At the beginning of the project, a larger number of services was used, which
provided more fine-granular analysis steps. In the course of the project, how-
ever, it turned out that a quite coarse-granular service library, which provides
predefined variants of the major analysis steps, rather than exposing computa-
tional details of the basic analysis steps to the workflow level, is more adequate.

5.2.2 FiatFlux Workflow for Single Data Sets

With these SIBs and the standard set of SIBs provided by Bio-jETI it was
now possible to define appropriate FiatFlux workflows. Figure 5.6 shows a
workflow for the analysis of a single data set, analogous to the abstract work-
flow depicted in Figure 5.5: The input CDF file is read and converted into
the FiatFlux format, and then used for flux ratio estimation and calculation
of net fluxes. In between the respective results are saved into text files for
later use.