Conclusion - User-Level Workflow Design: A Bioinformatics Perspective - page 193

Promotion
A central question with regard to making third-party users use the framework
is where and how in silico researchers would use semantics-based, automatic
workflow composition functionality. Larger service collections like EMBOSS
(cf. Section 3.3) or the BioCatalogue [41], which provide hundreds or even thou-
sands of services, are in fact not manageable without systematic discovery or
service composition techniques. When dealing with small domains, human ex-
perts may be unbeatable in composing tailoredworkflows,but not everyhuman
is an expert in bioinformatics services and data types, and also experts cannot
always keep track of all changes in large domain libraries. Thus both experts
and average users profit from tools that systematically collect and exploit se-
mantic service annotations and appropriately formalized domain knowledge.
However, although there are many promising approaches to semantics-
based, (semi-) automatic service composition, they have apparently not be-
come widely established so far. One reason for this might be that biologists are
reluctant to experiment with the new technologies, while bioinformaticians
have accustomed themselves to scripting and conventional programming in a
multitude of languages. Or, as stated by Ian Holmes in the BioWiki, “Bioin-
formatics workflows can be approached in several ways. Unix hackers often
resort to GNU make, while computer scientists dream of more elegant ap-
proaches.” [131]. Thus, computer science will have to explore further on how
to make their more elegant and more powerful approaches to (semi-) auto-
matic workflow design accessible to the broad biological user community.
The usability of concrete implementations of (semi-) automatic workflow
composition in Bio-jETI (and other systems) could, for instance, be assessed
by an evaluation framework as applied in [313, 244]. This would mean to
carry out the usability evaluation based on analysis of actually created work-
flows, user questionnaires and semi-structured interviews. However, the mere
technical aspects of usability are not the only aspects that are relevant for
the success of a particular software system or methodology. In fact, it also
largely depends on “the realms of sociology and psychology” [128], which
are sometimes dicult to understand. Hence, it would be interesting to ex-
plore potential users' obstacles for using semantic systems more generally. It
is likely that people simply mistrust the formal and/or semantic methods
that are still very unfamiliar to them. Here, specific promotion and training
(educational material, comprehensive tutorials, etc.) is required to get people
accustomed to a new way of thinking about workflows.
Furthermore, experience shows that an entire software design method is
dicult to explain and promote, since application experts tend to focus on
the aspects related to their specific domain, expecting a concrete software
tool for a very particular purpose. Thus, it seems to be a promising strategy
to carefully prepare domain models for specific bioinformatics applications,
and then release Bio-jETI together with a particular domain model as a
custom-tailored, special-purpose workflow tool.