Performance of the probe-set method in classifying incomplete genomes. Using the taxonomy database provided by NCBI, three organisms were randomly selected from each class to perform this experiment. A native probe-set classification tree of the 87 selected organisms was first constructed. Then, the genomes of these organisms were randomly truncated in different extents. With the truncated genomes, new classification trees were constructed and compared to the native tree. (a) Accuracy for the reclassification of a randomly truncated genome. The 87 genomes were subjected to random truncation and tree reconstruction one genome at a time. The accuracy (the y-axis) is computed as the ratio of truncated genomes that did not affect the topology of the tree. At each truncation rate, as indicated by the x-axis, the experiment was repeated 10 times and the standard deviation was thus obtained. (b) Similarity between the native classification tree and the tree with large-scale random truncations. All the 87 genomes were subjected to random truncation at the same time in this experiment. After a random truncation, the tree was reconstructed and its correlation coefficient (y-axis) with the native tree is computed to measure its topological similarity to the native tree. At each truncation rate, as indicated by the x-axis, the experiment was repeated 10 times and the standard deviation was thus obtained.
Lee et al. BMC Genomics 2012 13:157 doi:10.1186/1471-2164-13-157