AKID

AKID identifies kinase-specific phosphorylation events at single-kinase resolution across eukaryotic proteomes, predicting kinase–substrate interactions and inferring specificity-determining residues.

Key Features:

• Deep Neural Network Architecture: Employs a deep neural network trained on annotated human kinase–substrate (KsP) datasets to model kinase–substrate recognition.
• Automated Kinase Domain Detection: Automatically detects kinase domains within protein sequences across proteomes for mapping kinase activity.
• Residue Specificity Inference: Infers residues that contribute to kinase target specificity to identify specificity-determining positions.
• Target Peptide Prediction: Predicts potential kinase target peptides by applying learned sequence patterns from human KsP data to other eukaryotes.
• Cross-Species Application: Transfers specificity models derived from human KsP datasets to predict kinase–substrate interactions in diverse eukaryotic organisms.
• Conservation Analysis: Enables analysis of kinase specificity conservation among eukaryotes, including the reported high conservation of tyrosine kinase specificity.

Scientific Applications:

• Phosphorylation Network Mapping: Identifies novel kinase–substrate phosphorylation events to expand maps of cellular signaling networks.
• Clinical Relevance: Supports discovery of potential biomarkers and therapeutic targets, including the identification of over 1,590 novel kinase-specific phosphorylations in the human proteome.
• Research on Less-Studied Organisms: Predicts KsP events without requiring prior functional annotation, facilitating kinase–substrate discovery in poorly annotated eukaryotes.

Methodology:

AKID trains a deep neural network on annotated human KsP datasets, integrates known human phosphorylation data with the model, automatically detects kinase domains within proteomes, infers specificity-determining residues, and predicts target peptides based on learned sequence patterns.