:use 20 rounds of MICE for matrix completion

group predictions by source peptides before scoring their quality, added logging of each hyperparameter's performance

using proper cross-validation to estimate accuracy of just synthetic data (previously small leakage of information via shared similarity matrix)