DeepKhib

DeepKhib predicts lysine 2-hydroxyisobutyrylation (Khib) sites in protein sequences to identify potential post-translational modification locations involved in gene transcription and signal transduction.

Key Features:

• Deep Learning Algorithm: Employs a convolutional neural network (CNN) with one-hot encoding to learn sequence patterns indicative of Khib sites.
• Species-Specific and General Models: Implements species-specific models and a general model that integrates data from multiple species for cross-species prediction.
• Performance Metrics: Reports area under the ROC curve (AUC) ranges of 0.82–0.87 for species-specific predictions and 0.79–0.87 for general predictions, outperforming traditional machine-learning algorithms and other deep-learning models.

Scientific Applications:

• Regulatory Mechanism Analysis: Supports investigation of how Khib influences gene transcription and signal transduction regulatory mechanisms.
• Cross-Species Analysis: Enables comparative and evolutionary studies by predicting Khib sites across different species.
• Experimental Guidance: Prioritizes candidate Khib sites to guide experimental validation efforts.

Methodology:

Uses a convolutional neural network (CNN) with one-hot encoding; constructs species-specific and a general model by integrating multi-species data; evaluates predictive performance using AUC and comparisons with traditional machine-learning and other deep-learning models.