MusiteDeep

MusiteDeep predicts protein post-translational modification (PTM) sites from amino-acid sequences using deep learning to support analyses of PTM-dependent regulation, signaling, and disease mechanisms.

Key Features:

• Real-time prediction: Processes up to 1000 protein sequences per PTM type in under three minutes.
• Multiple PTM types: Performs simultaneous prediction of multiple PTMs to enable investigation of PTM cross-talk.
• Homology-based mapping: Maps predicted PTM sites to protein 3D structures via homology-based search.
• Comprehensive database: Maintains a local database of pre-processed PTM annotations from UniProt/Swiss-Prot refreshed every three months.
• Benchmarked performance: Rigorously benchmarked and shows competitive performance compared to other PTM prediction tools.
• Python tools for local computation: Provides Python tools for local computation.

Scientific Applications:

• Proteomics and functional analysis: Identifies modification sites that influence protein function and interactions.
• Regulatory network and PTM cross-talk analysis: Simultaneous prediction of multiple PTMs supports study of PTM cross-talk and regulatory networks.
• Signaling and disease mechanism investigation: Aids analysis of cellular signaling pathways and disease mechanisms by mapping PTM sites.

Methodology:

Employs deep-learning models that use only protein sequences as input without explicit feature extraction, applies ensemble techniques to improve accuracy, and maps predicted sites to 3D structures via homology-based search.