Figure 4

Comparison of performance of HGTector and conventional BLAST-based method on simulated genomes. The methods were tested on the simulated genomic data under an idealized (A) or randomized (B) tree topology. “Con” and “Rel” represent conservative and relaxed cutoffs in the HGTector analysis. “C = 0” and “D > C” are two criteria in the conventional BLAST-based method. Each experimental group is composed of 100 tests. The distribution of results in terms of precision (red) and recall (blue) is depicted by box plots. The mean value of each group is label above the corresponding column.