Deep-Kcr

Deep-Kcr predicts lysine crotonylation (Kcr) sites in proteins using a convolutional neural network to enable high-throughput proteomic analysis of this posttranslational modification.

Key Features:

• Convolutional Neural Network (CNN): employs a CNN architecture for Kcr site prediction.
• Feature integration: integrates sequence-based features, physicochemical property-based features, and numerical space-derived information.
• Feature selection: applies information gain for selecting informative features.
• Model evaluation: assessed using 10-fold cross-validation and independent set testing.
• Benchmarking: compared against long short-term memory networks (LSTM), random forests, LogitBoost, naive Bayes, and logistic regression.
• High-throughput prediction: enables large-scale computational screening of Kcr sites across proteins and species.

Scientific Applications:

• Kcr site identification: prediction of lysine crotonylation sites in proteins for proteomic studies.
• Posttranslational modification research: support for large-scale analyses of protein posttranslational modifications, including histone lysine crotonylation.
• Biological investigation: facilitating studies of the roles of histone lysine crotonylation in cellular regulation and disease-related processes.

Methodology:

Integrates sequence-based, physicochemical, and numerical space-derived features; selects features via information gain; trains a convolutional neural network; evaluates performance with 10-fold cross-validation and independent set testing and compares to LSTM, random forest, LogitBoost, naive Bayes, and logistic regression.