PROSPECT

PROSPECT predicts proteome-wide histidine phosphorylation substrates and sites from protein sequence to support analysis of cellular signaling pathways and metabolic regulation.

Key Features:

• Hybrid ensemble architecture: Integrates outputs from two convolutional neural network (CNN)-based classifiers with a random forest classifier.
• One-of-K Coding: Encodes categorical amino acid information as sparse binary vectors for model input.
• Enhanced Grouped Amino Acids Content (EGAAC): Represents grouped amino acid composition to capture local sequence properties.
• Composition of K-Spaced Amino Acid Group Pairs (CKSAAGP): Encodes k-spaced amino acid group pair composition to capture spatial arrangement within sequences.
• Classifier-feature mapping: Uses each of the three feature sets as input to one of the classifiers—two CNNs and one random forest—to generate complementary predictions.
• Sequence-based proteome-wide prediction: Predicts both histidine phosphorylation substrates and specific phosphorylation sites from primary sequence data.

Scientific Applications:

• Cellular signaling pathways: Enables identification of histidine phosphorylation sites relevant to signal transduction studies.
• Metabolic processes: Supports investigation of histidine phosphorylation roles in metabolic regulation.
• Cross-organism analysis: Applicable to prokaryotic proteomes and provides potential insights into analogous mechanisms in mammalian cells.
• Protein function and regulation: Contributes sequence-based evidence for studies of protein regulation mediated by histidine phosphorylation.

Methodology:

Applies One-of-K Coding, EGAAC, and CKSAAGP feature encodings as inputs to two CNN-based classifiers and one random forest classifier, and integrates their outputs via a hybrid ensemble.