SNN6mA: Improved DNA N6-methyladenine site prediction using Siamese network-based feature embedding

Area of Science:

• Genomics
• Epigenetics
• Bioinformatics

Background:

• DNA N6-methyladenine (6mA) is a crucial epigenetic modification involved in diverse biological processes.
• Accurate identification of 6mA sites is vital for understanding its regulatory roles and functions.
• Existing DNA 6mA site prediction methods require improvement for enhanced accuracy.

Purpose of the Study:

• To develop a highly accurate predictor for DNA 6mA sites using a Siamese network.
• To improve the discriminative feature extraction for 6mA site prediction in DNA sequences.

Main Methods:

• DNA segments were encoded using one-hot encoding.
• A Siamese network was employed to map feature vectors into a low-dimensional embedding space.
• A fully connected neural network processed the learned features for final prediction.

Main Results:

• The proposed SNN6mA predictor demonstrated superior performance compared to state-of-the-art methods on benchmark datasets from two species.
• The Siamese network effectively captured more discriminative features in a low-dimensional space.
• SNN6mA achieved higher accuracy in predicting DNA 6mA sites.

Conclusions:

• SNN6mA represents the first application of Siamese networks for DNA 6mA site prediction.
• The method offers improved accuracy and discriminative feature learning for epigenetic modification identification.
• The SNN6mA approach is adaptable for predicting other DNA modifications.