DeepCap-Kcr

DeepCap-Kcr predicts lysine crotonylation (Kcr) sites on histone and nonhistone proteins using a deep learning architecture to support analysis of protein posttranslational modification relevant to chromatin remodeling, cell cycle regulation, and proteomic diversity.

Key Features:

• Lysine crotonylation (Kcr) site prediction: Identifies Kcr sites on both histone and nonhistone proteins.
• Deep learning architecture: Integrates a capsule network (CapsNet) with convolutional neural networks (CNNs) and long short-term memory (LSTM) units.
• Internal capsule layer motif detection: Uses the internal capsule layer to explore data distributions and detect motifs with biological significance.
• Hierarchical representation learning: Captures hierarchical representations and multi-level features from sequence data.
• Cross-species applicability: Generalizes across species, including mammalian proteins and papaya proteins.
• Robustness to limited samples: Architecture is reported to be beneficial when training data are limited.
• Performance versus existing models: Outperforms CNN-based models such as Deep-Kcr.

Scientific Applications:

• Cell cycle regulation studies: Predicts Kcr sites to investigate proteins involved in cell cycle regulation.
• Cell organization research: Maps Kcr modifications to study roles in cell organization.
• Chromatin remodeling investigations: Identifies histone Kcr sites relevant to chromatin remodeling.
• Proteomic diversity analyses: Expands analyses of proteomic diversity by locating Kcr on nonhistone proteins.
• Disease mechanism and progression research: Supports study of human disease progression through Kcr site prediction.
• Drug development: Reveals Kcr-related motifs and functional sites that can inform drug development.
• Basic proteomics research: Enables investigation of protein posttranslational modifications in basic research contexts.

Methodology:

The model integrates a capsule network (CapsNet) with convolutional neural networks (CNNs) and long short-term memory (LSTM) units, uses the internal capsule layer to explore data distributions and detect motifs, and captures hierarchical representations and multi-level features from sequence data.