saturating mutagenesis wild-type residue

[SkwisgaarSkwigelf]

Just a general question. I'm running single site saturating mutagenesis as per the protocol using run_example_2.py as a template using the suggested parameters:

nstruct = 35 # Normally 35
max_minimization_iter = 5000 # Normally 5000
abs_score_convergence_thresh = 1.0 # Normally 1.0
number_backrub_trials = 35000 # Normally 35000
backrub_trajectory_stride = 35000 # Can be whatever you want, if you would like to see results from earlier time points in the backrub trajectory. 7000 is a reasonable number, to give you three checkpoints for a 35000 step run, but you could also set it to 35000 for quickest run time (as the final minimization and packing steps will only need to be run one time).

The script runs fine, and when I analyze the output I noticed that the wild-type amino acid is included among the sampled mutations. Not surprising given that I didn't alter this line:

for mut_aa in 'ACDEFGHIKLMNPQRSTVWY':

However, occasionally I see that the wild-type amino acid scores fairly well (~ -0.5 to -1.0 kcal/mol) in both the fa_talaris2014 and fa_talaris2014-gam ddG metrics. Obviously, I would expect the ddG value of a wt-to-wt substitution to be 0, so I'm wondering if this large of a discrepancy is indicative of the amount of error inherent in the ddG calculations? I worry if other predicted stabilizing mutations are real given that the wild-type residue scores so well. Has this ever been seen before?