Fig. 8
Two data augmentation methods. In (a), R is reference sequence. B is the HGT insertion site, which is in the middle of training sequence s. s1 and s2 are augmented training samples, which are sampled from near the HGT insertion site B within a small region. In (b), through randomly changing nucleotides in s, we get the augmented training sample s′, which retains most nucleotides of s