i6mA-Caps

i6mA-Caps identifies DNA N6-methyladenine (6mA) sites using a Capsule Network (CapsNet) to predict epigenetic modification locations for studies of gene regulation.

Key Features:

• Capsule Network (CapsNet) Framework: Leverages a Capsule Network architecture to capture spatial hierarchies in sequence-derived features beyond traditional convolutional neural networks.
• Single Encoding Scheme: Employs a numerical representation scheme for DNA sequences to enable computational processing.
• Convolution Layers: Uses convolutional layers to extract low-level features from the encoded numerical DNA data.
• Capsule Layers: Processes convolutional features with capsule layers to derive intermediate and high-level representations for classification.
• Reported Accuracy: Demonstrated accuracies of 96.71% on the Rosaceae dataset, 94% on the Rice dataset, and 86.83% on the Arabidopsis thaliana dataset.
• Comparative Performance: Reportedly outperforms existing state-of-the-art methods for 6mA site identification.

Scientific Applications:

• Epigenetic Research: Enables identification of 6mA sites to support studies of epigenetic mechanisms and regulation of gene expression.
• Genomic Studies: Facilitates exploration of the distribution and biological roles of 6mA across diverse genomes such as Rosaceae, rice, and Arabidopsis thaliana.

Methodology:

DNA sequences are encoded numerically, features are extracted via convolutional layers, and the resulting features are processed by a capsule network for classification of 6mA sites.