Below is a graph of each of the unperturbed data sets from different scoring matrices. A general observation that can be made is that the PAM250 Matrix, which scored the poorest in terms of minimum epsilon value, also has the poorest performance in our experiment.
The next graph is a comparison of several perturbed and unperturbed versions of BLOSUM62. Again, the one with the poorest result from CPLEX (in other words, the highest minimum epsilon) performed the poorest in our experiment.
