SNO-DCA

SNO-DCA predicts S-nitrosylation sites in protein sequences using deep learning to identify residues modified by S-nitrosylation for analysis of this post-translational modification.

Key Features:

• Deep Learning Architecture: Employs densely connected convolutional blocks (Dense Convolutional Blocks) combined with an attention mechanism to perform site-level prediction.
• One-Hot Encoding: Transforms protein sequences into one-hot encoded representations as the model input.
• Feature Extraction and Attention Mechanism: Uses Dense Convolutional Blocks to extract sequence features and an attention module to assign weights to informative features.
• Performance Under Imbalanced Data: Demonstrates superior performance under imbalanced datasets, validated by 10-fold cross-validation and independent testing compared to existing models.

Scientific Applications:

• Mapping S-nitrosylation Sites: Predicts candidate S-nitrosylated residues to support studies of protein regulation and function.
• High-Throughput Annotation: Provides computational predictions as an alternative or complement to experimental identification of S-nitrosylated sites.
• Drug Discovery and Disease Research: Supports investigation of diseases associated with aberrant S-nitrosylation and aids prioritization of targets for drug discovery.

Methodology:

Sequence one-hot encoding, feature extraction using Dense Convolutional Blocks, integration of an attention mechanism, and validation via 10-fold cross-validation and independent testing.