Table 2.

Number of misclassifications for two temperature optimum ranges, hyperthermophiles versus non-hyperthermophiles (hyper vs. non-hyper) and mesophiles versus non-mesophiles (meso vs. non-meso), using all tetramer frequencies over the entire virtual coding strand. Model-with-all is the number of misclassifications when building a model using all 195 genomes. Leave-one-out is the number of mispredictions when classifying an organism against a model built using the other 194 genomes. 50-randomly-selected-genomes is the average number (over 40 iterations) of misclassifications when building a model on 145 genomes and attempting to classify the other 50 randomly selected genomes.

Temperature range	Model-with-all	Leave-one-out	50-randomly-selected-genomes
Hyper vs. non-hyper	2 (1%)	8 (4.1%)	2.6 (5.2%)
Meso vs. non-meso	9 (4.6%)	14 (7.2%)	5.1 (10.2%)