Information Technology Reference
In-Depth Information
As the tools required for the second example are also part of the EMBOSS
suite, they are available as processors in the same way. Similar to the first
example the workflow for constructing a phylogenetic tree (see Figure 8.9)
starts with sending a set of sequences to
emma
. The resulting alignment is then
forwarded to
fseqboot
, which generates a set of reference sequences. Now the
continuation of the analysis depends on the type of the input data: nucleic
acid sequences have to be processed by
fdnapars
, amino acid sequences by
fprotpars
.
The conditional branching that is required here can be realized in Tav-
erna by using control links and an additional (boolean) input specifying if
the input is DNA or not. This boolean value is (data-) linked to two pro-
cessors,
fail if true
and
fail if false
, which are (control-) linked (the
gray connectors with the circle in Figure 8.9) to
fprotpars
and
fdnapars
,re-
spectively. Only one of
fail if true
and
fail if false
can be successfully
passed during execution, and subsequently either
fprotpars
or
fdnapars
has both input links available and will be executed. The remainder of the
workflow is straightforward: the resulting phylogeny is directly used as out-
put and additionally sent to
fconsense
in order to compute the consensus
tree from the large number of possible results.
Example 3 in Taverna
Although the DDBJ web services on which this example is based can be
accessed via Taverna processors, it is not possible to realize the whole process
in the workbench. The reason is that the aggregation of information for the
resulting table takes place in the inner of two nested for-style loops: The
outer loop iterates over the list of accession numbers that are returned by
ARSA, and the inner loop processes the results of the BLAST search that is
performed for each accession number.
While this so-called parameter sweeping can be built in control-flow-based
systems in a straightforward manner, it has not been possible to implement
it in Taverna's data-flow-oriented environment when using only the function-
ality that is readily provided by the workbench. The implicit loops that are
usually provided by the latter are sucient for batch-like processing of data-
sets, but fail when the well-directed addressing of data items within the loop
is necessary.
Search WWH ::
Custom Search