Our earlier studies confirmed the usefulness
of this approach, but also found that BPAnalysis,
the implementation developed by Sankoff and Blanchette, was too slow to
use on all but very small datasets. We report here on a reimplementation
of BPAnalysis using the principles of algorithmic engineering.
Our faster (by 2 to 3 orders of magnitude) and flexible implementation
allowed us to conduct studies on the characteristics of breakpoint analysis,
in terms of running time, quality, and robustness, as well as to analyze
datasets that had so far been considered out of reach. We report on these
findings and also discuss future directions for our new implementation.