Figure 1. Schematic overview of the two methods used to predict oxygen requirement in bacteria.

a) In the one-step prediction method the genomes in the test set are assigned a posterior probability for each of the three included classifications, given their protein domain profile. The genomes are predicted to belong to the classification to which they have the highest posterior probability. b) The genomes in the test set are first assigned posterior probabilities for being able or unable to respire, based on their protein domain profile. Using a second model, those genomes found most likely to be capable of respiration are assigned a posterior probability of belonging to the classifications Aerobe or Facultative.