Metagenome Fragment Classification Using 𝑁-Mer Frequency Profiles

Figure 10

5-fold cross-validation performance for species-accuracy on 9 species classes (77 strains), training on ∼62 strains and testing on ∼15 strains at a time. The maximum performance for each fragment size is 97.3%±1.0% for 500 bp, 95.3%±1.3% for 100 bp, and 90.2%±1.2% for 25 bp fragments, demonstrating that the NBC has good classification species accuracy for never seen before strains.