Related Work - User-Level Workflow Design: A Bioinformatics Perspective - page 173

8.2.1 Reference Workflows

For the comparison of Taverna and Bio-jETI with respect to their workflow
modeling capabilities three increasingly complex reference workflows have
been selected. They are introduced in the following, before their realizations
in both workflow systems are discussed. All examples are third-party work-
flows in order to make sure that the characteristics of the benchmarks have
not been influenced by the system under consideration beforehand. In particu-
lar, examples from the myExperiment workflow repository [113], for instance,
which primarily contains Taverna workflows, would not have been suitable
for this purpose.

Workflow Example 1: Multiple Sequences Analysis

consensus
output

cons

sequence
input

emma

sequence
output

ehmmbuild ehmmcalibrate ehmmemit

Fig. 8.1 Multiple sequences analysis workflow

The first example workflow (sketched in Figure 8.1) originates from the We-
bLab project [349]. WebLab is a web-based platform for executing and com-
bining bioinformatics analysis tools, predominantly of the EMBOSS suite (cf.
Section 3.3). Therefore, this example is mainly based on EMBOSS tools. The
workflow is basically linear, but as the figure shows can easily be parallelized, as
intermediate results are used independently by different subsequent services.
Taking a set of (nucleic or amino acid) sequences as input, the workflow
computes their consensus sequence and generates set of corresponding new
sequences. Therefor it starts with a multiple sequence alignment computed
by an EMBOSS program called emma , which is an interface to the ClustalW
algorithm. A consensus sequence is calculated from the alignment using cons .
The alignment is also used as input for ehmmbuild , which builds a profile Hid-
den Markov Model (HMM) from the alignment. The HMM search statistics
are calibrated by ehmmcalibrate ,before ehmmemit is used to generate a set
of sequences from the HMM.

Workflow Example 2: Phylogenetic Tree Construction

protein

fprotpars

sequence
input

tree
output

emma fseqboot fconsense

DNA

fdnapars

Fig. 8.2 Phylogenetic tree construction workflow