Introduction - User-Level Workflow Design: A Bioinformatics Perspective - page 27

[353]. It integrates well into the entire data analysis process, starting from
the raw data sets and spreadsheet experiment documentations from the
lab, and producing outputs that can be processed by a specific visualiza-
tion software.
As there is only a small number of actual analysis services involved, the
variants of the FiatFlux-P workflows are mainly imposed by the possible
variations in data handling. Nevertheless, these variants are numerous
and often run off the actual track, so that for synthesizing FiatFlux-P
workflows comparatively many constraints have to be provided to take
the relevant dependencies into account and characterize the intended so-
lution(s) appropriately.

Application scenario 4: microarray data analysis workflows
The microarray data analysis workflows provide a framework for user-level
construction of analysis pipelines based on appropriately wrapped and in-
tegrated Bioconductor [105] functionality. It helps handling the variability
that is inherent in microarray data analysis at an user-accessible level and
thus again emphasizes the agility of the model-driven and service-oriented
approach to workflow design.
Remarkably, the workflows in this scenario comprise comparatively
many analysis steps, but typically have a simple, in fact mostly linear,
structure. Thus, the automatic workflow composition functionality of Bio-
jETI, which is based on a linear-time logic synthesis algorithm, can be
applied here to generate the complete analysis pipelines automatically.
Although the applications address different scenarios with different char-
acteristics and make use of highly heterogeneous services, they are still real-
izable homogeneously within the Bio-jETI framework. By using a coherent
workflow formalism, it becomes possible to focus on the application-level
aspects of the different workflows. In fact, the gained insights about the
(technical, service-level) characteristics of the application domain(s) - system-
atically and adequately captured and formalized - enhanced the knowledge
base required for the successful application of semantic techniques to support
workflow design.
Furthermore, the different applications provide a proper basis for further
research and considerations about bioinformatics workflows in particular and
scientific workflows in general. Most importantly, in this work they are used
as basis for the first systematic analysis and evaluation of the central syn-
thesis method with regard to performance and the impact of domain model
and constraint patterns. To this aim, an evaluation framework has been con-
ceived for analyzing the impact of constraints on results and performance
systematically.
On the one hand, the evaluation shows that constraints can directly in-
fluence the synthesis results by guiding the search to the actually intended
solutions. On the other hand, it shows that constraints can also have a posi-
tive effect on the synthesis performance by decreasing the size of the search
space considerably. This is particularly relevant since state explosion, that is,

•