RattnetFold

RattnetFold predicts protein fold types from protein sequences and residue-residue contact maps using a stack convolutional neural network with attention mechanisms and metric learning to extract fold-specific features.

Key Features:

• Stack Convolutional Neural Network: Employs a stack convolutional neural network (CNN) to extract hierarchical features and capture intricate patterns within residue-residue contact maps.
• Attention Mechanism: Integrates an attention mechanism to emphasize significant regions within protein sequences and contact maps for improved feature extraction.
• Sequence-based Feature Extraction: Utilizes a sequence-based approach to derive representative features from protein sequences.
• Metric Learning with RattnetFoldPro: Applies metric learning (RattnetFoldPro) to project fold-specific features into a subspace where proteins with similar folds are positioned closer together to enhance discriminative power.

Scientific Applications:

• Protein Fold Prediction: Improves accuracy of protein fold recognition and classification from sequence and contact map data.
• Functional Annotation: Supports inference of protein function and interactions via fold-specific structural features.
• Drug Discovery: Informs structure-based drug discovery efforts by providing more accurate fold assignments.
• Disease Research: Aids investigation of disease-related proteins by clarifying structural fold relationships.
• Biomaterials Development: Assists design and development of novel biomaterials through enhanced understanding of protein folds.

Methodology:

Processes protein residue-residue contact maps and sequences using a stack convolutional neural network augmented with an attention mechanism, and employs metric learning in RattnetFoldPro to project fold-specific features into a discriminative subspace.