This document summarizes a novel statistical method for discriminating between thermostable proteins. The method uses amino acid frequency, dipeptide frequency, and physical-chemical features extracted from protein sequences to classify proteins as mesophilic, thermophilic, or hyper-thermophilic. The researchers tested the effect of each set of features individually and combined on classification accuracy. Their results show the proposed method using all feature types together achieved the best performance, correctly classifying over 90% of proteins on average.