Phylogenetic Analysis Workflows - User-Level Workflow Design: A Bioinformatics Perspective - page 84

This constraint decreases the number of solutions further down to 169 in
depth 2, then 7,933 in depth 3, and 274,664 in depth 4 (see row 3 of Table
3.2).
Similarly, it can be observed that often multiple sequence alignment ser-
vices (like, e.g., emma ) are used, which produce phylogenetic trees as more
ore less by-products of the alignment computation. If it is not desired to
derive phylogenetic trees this way, also Sequence alignment services can be
excluded from the solutions:

• Do not use Sequence alignment .

With this constraint, no one-step solutions are found any more and there
is only a single two-step solution (see row 4 of Table 3.2). Furthermore, the
number of solutions found until a depth of 3 is decreased further to 132, and
to 8,801 until a depth of 4.

Refinement 3: Expression of Intents

Still, this number of solutions is not really manageable for the user. In order to
steer the search towards the intended solutions more e ciently, the synthesis
algorithm can be provided with constraints that express specific ideas about
the desired workflow.
For instance, the service enforcement constraint can be used to explicitly
demand the inclusion of a phylogenetic tree construction service that uses a
parsimony method:

• Enforce the use of Phylogenetic tree construction (parsimony methods) .

With this constraint, first solutions are found in depth 3, and the synthesis
returns a total of 271 solutions in depth 4 (see row 5 of Table 3.2).
If the user even has concrete tools in mind that he would like to see in the
solution, he can enforce their existence. For instance, if the input nucleotide
sequence should be translated into a protein sequence by transeq and then
a phylogenetic tree be created using fprotpars , corresponding constraints
can be applied:

• Enforce the use of transeq .

• Enforce the use of fprotpars .

These constraints narrows the search to a single solution of length 3 and
a still assessable set of 134 solutions in depth 4 (see row 6 of Table 3.2).

Refinement 4: Alternative Synthesis Strategy

Another, general observation about the obtained solution is that often work-
flow parts without proper relationships to the surrounding workflow are cre-
ated. Most prominently, outputs of services are often not adequately used