I-DNAN6mA: Accurate Identification of DNA N⁶-Methyladenine Sites Using the Base-Pairing Map and Deep Learning

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• DNA N⁶-methyladenine (6mA) is crucial in biological processes, but its roles are not fully understood.
• Existing experimental methods for identifying 6mA sites are slow and costly, hindering research progress.

Purpose of the Study:

• To develop a novel, rapid, and cost-effective computational method for identifying 6mA sites.
• To complement existing experimental techniques and accelerate the study of 6mA functions.

Main Methods:

• Proposed I-DNAN6mA, a computational tool utilizing base-pairing rules and a three-stage deep learning model with pairwise inputs.
• Developed a unique image-like representation for DNA sequences.
• Evaluated performance across four species: *Arabidopsis thaliana*, *Drosophila melanogaster*, *Rice*, and *Rosaceae*.

Main Results:

• I-DNAN6mA demonstrated high performance with Area Under the Curve (AUC) values ranging from 0.947 to 0.990.
• Achieved accuracies between 88.2% and 96.2% and Mathew's Correlation Coefficient (MCC) values from 0.763 to 0.924.
• Outperformed several existing state-of-the-art methods in identifying 6mA sites.

Conclusions:

• I-DNAN6mA is the first method to use image-like DNA representation and pairwise deep learning inputs for 6mA site identification.
• The method is expected to significantly aid in locating functional DNA regions and advancing 6mA research.