CapsNh-Kcr

CapsNh-Kcr predicts lysine crotonylation (Kcr) sites on non-histone proteins using a capsule network-based deep learning model to improve post-translational modification site identification and characterize motif distributions.

Key Features:

• Capsule network architecture: Employs capsule networks as the deep learning architecture for site prediction.
• Non-histone specificity: Specifically focuses on predicting lysine crotonylation (Kcr) sites in non-histone proteins.
• Motif discovery: Identifies and uncovers significant motif distributions across predicted Kcr sites.
• Performance: Achieves an area under the curve (AUC) of 0.9120, approximately 6% higher than previous models focused solely on non-histone proteins.
• PTM context: Targets lysine crotonylation (Kcr), a post-translational modification implicated in metabolism and cell differentiation.

Scientific Applications:

• Candidate site prioritization: Prioritizes candidate Kcr sites in non-histone proteins for experimental validation.
• Motif pattern analysis: Maps motif patterns in non-histone Kcr sites to support studies of Kcr-mediated regulation.
• Biological investigation: Supports investigation of Kcr roles in biological processes such as metabolism and cell differentiation.

Methodology:

Implements a capsule network-based deep learning model to predict Kcr sites and analyze motif distributions, with performance evaluated by area under the curve (AUC = 0.9120) and compared to previous non-histone models (~6% improvement).