The first is the inadequate learning of the imbalanced m6A samples which are much less than the non-m6A samples, by their balanced learning approaches. Second, the features used by these methods are not outstanding to represent m6A sequence characteristics. |