FiatFlux-P - User-Level Workflow Design: A Bioinformatics Perspective - page 105

square optimization. In case the system is underdetermined, netto can either
estimate all calculable fluxes or optimize the solution with regard to a partic-
ular flux or intermediate. In any case, the user has to start the computation
by pressing one of the buttons in the “Analysis” frame of the netto window.
When the calculation has finished, an F value that indicates the quality of the
obtained solution is displayed within the frame. If the user considers the F value
being too large, he can either modify the network model and restart the calcu-
lation, or simply re-run the calculation with the same network configuration,
since the computations are partly based on random numbers and can thus lead
to different results also without parameter changes.
Summarizing, FiatFlux is a highly interactive desktop software that pro-
vides a graphical user interface to a Matlab implementation of METAFoR
and 13 C-based metabolic flux analysis. For an in-depth explanation of the
software and the calculations the reader is referred to [95] and [230].

5.2 FiatFlux-P Workflows

Naturally, interactive workflows become limiting when hundreds or even thou-
sands of data sets have to be handled. The previous chapter showed how the
highly interactive GeneFisher web application has been turned into a flexible
workflow realization, called GeneFisher-P. Metabolic flux analysis with Fiat-
Flux is a similarly interactive process that has not been designed for batch
operation. Like GeneFisher-P, FiatFlux-P [90] is a workflow-based version of
the original interactive analysis software that allows the user to work with
the FiatFlux functionality in a highly flexible and automated manner. It has
been developed in the scope of a joint effort with the Laboratory of Chemi-
cal Biotechnology at the TU Dortmund, where it has actively been used for
large-scale data analyses (cf. [89]).
Among the Bio-jETI applications, FiatFlux-P is the application where the
speedup of the analyses was most considerable: the analysis with FiatFlux-P
runs about three to five times faster than the corresponding manual analysis.
Thatis, a complete 13 C-basedmetabolic flux analysis performed with FiatFlux-
P requiresabout four minutes instead of 12to 20minutes needed for the manual
analysis of a single data set. As metabolic flux experiments do not only produce
a single data set that has to be analyzed, but often 20, 50 or even 150 data
sets, this means that the time spent for the data analyses for an experiment
is now only about 1:20 h, 3:20 h, or 10 h instead of up to 6:40 h, 16:40 h, or
50 h, respectively. Furthermore the manual analysis requires the full attention
of an (experienced) human user, hence it is expensive in the sense that it can
easily consume a whole man-week of work. In contrast, the automatic analysis
workflowsrunautonomouslyin the background,possiblyalsoovernight,so that
the researcher can deliberately focus on other tasks in the meantime.