DeepEfflux

DeepEfflux identifies efflux protein families using a 2D Convolutional Neural Network (CNN) trained on Position-Specific Scoring Matrix (PSSM)-derived sequence features to support transporter biology research.

Key Features:

• 2D Convolutional Neural Network (CNN) Architecture: Employs a 2D CNN to capture sequence motifs surrounding hidden target residues and use these motifs as family-specific features.
• Position-Specific Scoring Matrix (PSSM) Input: Uses PSSMs to represent sequence conservation and variability as input features for the model.
• Independent Test Accuracies: Reports independent test accuracies of 96.02% for Class A, 94.89% for Class B, and 90.34% for Class C.
• Cross-Validation: Validated using 5-fold cross-validation across three distinct datasets, each representing a different efflux protein family.

Scientific Applications:

• Transporter biology studies: Identification of efflux protein families to support analyses of their structural and functional properties.
• Xenobiotic defense and energy dependence investigations: Elucidation of roles in cellular defense against xenobiotics and characterization of energy dependencies within efflux systems.

Methodology:

Training a 2D CNN on datasets of efflux protein families using PSSM-derived sequence features and sequence motifs, with evaluation by 5-fold cross-validation and independent test sets.