Introduction - User-Level Workflow Design: A Bioinformatics Perspective - page 26

its specific characteristics, and there are considerable differences between the
applications:
• Application scenario 1: phylogenetic analysis workflows

The first application scenario is concerned with phylogenetic analyses
based on molecular sequences, which have become a standard applica-
tion domain for illustrating bioinformatics workflow technologies (cf., e.g.,
[256, 134, 155, 172, 265, 165]). Accordingly, there is a plethora of easy-
to-use software tools available for the individual analysis steps, and the
workflows of this example could be built exclusively using publicly avail-
able services for the computations. This is in contrast to the other scenar-
ios, where specifically designed services were required for certain analysis
steps.
Not least due to the large number of available services, phylogenetic
analysis workflows are extremely variant-rich. Hence, they are well suited
to demonstrate the agility of workflow design in Bio-jETI and in partic-
ular the features and benefits of the constraint-driven, semi-automatic
workflow design approach. Furthermore, the annotation with semantic
meta-data is particularly advanced for this discipline [179], which is also
advantageous for the application of the constraint-driven workflow design
methodology.

• Application scenario 2: GeneFisher-P

The GeneFisher-P workflows [177] were developed in order to provide a
flexible alternative to the monolithic GeneFisher web application for PCR
primer design [109, 124]. The workflow version of the application makes
it easy, for example, to change individual steps (e.g. using alternative
sequence alignment algorithms), to increase or decrease the amount of
required user interaction (e.g. by letting the user assess each intermediate
result or by limiting user interaction to the specification of the input
data), or to define batch processing workflows that perform primer design
for several input sequences autonomously (rather than requiring manual
execution for every single input sequence).
The actual sequence of analysis services that has to be applied for per-
forming primer design depends on the specific type of the input data,
which is however not known at application modeling time. The origi-
nal application and workflow models thus include points of conditional
branching, where the user is asked to provide the required data character-
izations at runtime, and accordingly an appropriate sequence of services
is executed. Using the synthesis functionality of the new framework, the
user can generate the services sequence best suited for his concrete input
data ad hoc, enabling him to exhaust all possibilities provided by the
possibly evolving service libraries.

• Application scenario 3: FiatFlux-P

FiatFlux-P focuses on the automation of metabolic flux analysis. It is de-
signed to work off large sets of data from 13 C tracer experiments based on
a specialized version of the interactive FiatFlux desktop analysis software