GPS-Palm

GPS-Palm predicts S-palmitoylation sites on proteins to identify cysteine residues likely undergoing reversible lipid modification for experimental prioritization.

Key Features:

• Deep Learning Framework: Employs a convolutional neural network (CNN) architecture to analyze protein sequences for S-palmitoylation site prediction.
• Graphic Presentation System (GPS 6.0): Translates sequence similarity values into visual images to provide enriched feature representations for prediction.
• Data Quality Discrimination (DQD): Assigns data quality weights (DQWs) between different types of known S-palmitoylation sites derived from small- and large-scale experimental datasets.
• Integration of Multiple Features: Integrates nine sequence-based and structural features to enhance prediction accuracy across species.
• Parallel CNNs (pCNNs): Utilizes parallel CNN architectures to process multiple feature types simultaneously, improving robustness and accuracy.

Scientific Applications:

• General Prediction: Demonstrates superior general S-palmitoylation site prediction with an area under the curve (AUC) improvement of over 31.3% compared to existing tools.
• Species-Specific Predictors: Provides specialized predictors for human and mouse proteins with reported AUC values of 0.900 and 0.897, respectively.

Methodology:

Compiled a benchmark dataset of 3,098 known S-palmitoylation sites from experimental studies and trained CNN-based deep learning models, including parallel CNNs, integrating GPS 6.0 graphic representations, DQD-assigned data quality weights, and nine sequence-based and structural features.