Lessons Learned - User-Level Workflow Design: A Bioinformatics Perspective - page 140

Fig. 7.1 Synthesis statistics (number of solutions) for the EMBOSS scenario
for this case runs only until a search depth of 3, where the last complete
numbers have been obtained. Similarly, several lines run only until a search
depth of 4 or 5, which means that for these cases the maximum capacity of
the solution store has been reached in depth 5 or 6, respectively. Still, these
lines show the effect of the constraints on the number of solutions, enabling
the synthesis algorithm to proceed to greater search depths. In the last case
the solutions of length 6 can be fully explored, resulting in 8,317 remaining
solutions.
In total, the application of constraints and the use of particular synthesis
configurations tames the growth of solutions effectively: instead of exceeding
the solution store capacity already in a search depth of 4, the synthesis finally
returns a comparatively handy set of solutions of length less than or equal
to 6. Nevertheless the reached search depth is relatively small. Although fur-
ther constraints could restrain the solutions further and thus avoid exceeding
the capacity of the solution store also for greater search depths, eventually
the execution time of the synthesis exceeds feasible ranges. Consequently,
when working on the discussed application scenarios, the synthesis was simply
aborted when it became clear that it would be running “too long”, which typ-
ically meant running for days or weeks and not for minutes or hours any more.
(The next section discusses the performance of the synthesis in greater detail.)
Figure 7.2 visualizes until a search depth of 12 the number of synthesis solu-
tions that are obtained in the GeneFisher-P scenario (cf. Section 4.3) when no
constraints,the domainconstraints,andconstraintsfor the enforcement/avoid-
ance of DNA backtranslation and Sequence alignment are applied. In the uncon-
strained case, the number of solutions grows exponentially, returning 101,842
solutions in depth 12.The domainconstraintsreducethe number of solutions in
depth 12 to 1,086 solutions, and additionally cause that no solutions are found
before a search depth of 5. Enforcing the use of DNA backtranslation and Se-
quence alignment services yields that the shortest two solutions have a length