To create fashions for predicting auxotrophy from COG annotations, we skilled choice tree classifiers primarily based on the function choice course of offered above, or randomly chosen COG phrases from every pathway (500 randomizations). First, we appended the dataset of 35 validated strains with further 28 strains from the drop-out display screen offered in Supplementary Fig. 1 to lower sampling bias. This complete set of 63 was divided to balanced coaching and testing datasets utilizing 40% of the information for testing utilizing “train_test_split” and choice tree classifiers had been generated with “DecisionTreeClassifier” from sklearn library with a most depth of three nodes.
Comments
0 comments