DeepCys

DeepCys predicts multiple functional roles of cysteine residues in proteins using a deep neural network trained on protein crystallographic data to support structural functional annotation.

Key Features:

• Deep neural network-based prediction: Uses a deep learning framework trained on curated datasets derived from protein crystal structures to assign cysteine functions.
• Multiple function predictions: Predicts four distinct cysteine roles: disulfide bond formation, metal-binding, thioether linkage, and sulphenylation.
• Input requirements: Accepts PDB identifier (ID), chain ID, and residue ID for a target cysteine residue.
• Probabilistic output: Reports probabilities for each of the four predicted functions.
• Comprehensive analysis: Integrates local and global protein properties including sequence motifs, secondary structure elements, buried fractions, microenvironments, and protein/enzyme class information.
• Performance superiority: Demonstrates higher predictive power and broader scope compared with other multiple and specific cysteine function prediction algorithms.
• Thioether prediction novelty: Implements explicit prediction of thioether linkage functions.

Scientific Applications:

• Cysteine functional annotation: Assigns likely biochemical roles to cysteine residues to inform protein functional studies.
• Analysis of domains of unknown function: Aids characterization of proteins containing unannotated or poorly characterized domains by predicting cysteine roles.
• Case study application: Applied to analysis of cytochrome C oxidase subunit-II like transmembrane domains to uncover cysteine functional insights.

Methodology:

Trains a deep neural network on two independent datasets curated from protein crystallographic data and integrates diverse structural and contextual features for prediction.