Using ATCLSTM-Kcr to predict and generate the human lysine crotonylation database.

Area of Science:

• Biochemistry
• Genomics
• Computational Biology

Background:

• Lysine crotonylation (Kcr) is a vital post-translational modification involved in cellular processes like gene transcription and cancer.
• Predicting Kcr sites is essential for understanding its roles, but experimental methods are costly and time-consuming.
• Existing computational methods often lack accuracy and robustness in Kcr site prediction.

Purpose of the Study:

• To develop a highly accurate deep learning model for predicting lysine crotonylation sites.
• To create a benchmark dataset addressing mass spectrometry (MS) detectability limitations.
• To establish a comprehensive database for human Kcr site prediction and analysis.

Main Methods:

• Developed the ATCLSTM-Kcr model using a self-attention mechanism and natural language processing (NLP) techniques.
• Designed a pipeline to generate an MS-based benchmark dataset, mitigating false negatives.
• Integrated ATCLSTM-Kcr and other deep learning models to score human proteome lysine sites for the Human Lysine Crotonylation Database (HLCD).

Main Results:

• The ATCLSTM-Kcr model demonstrated superior accuracy and robustness compared to existing prediction tools.
• The developed MS-based dataset improved the sensitivity of Kcr prediction by addressing MS-detectability issues.
• The HLCD provides a comprehensive resource, scoring all human lysine sites and annotating experimentally identified Kcr sites.

Conclusions:

• ATCLSTM-Kcr offers an effective computational approach for Kcr site prediction, reducing experimental costs.
• The HLCD serves as a valuable integrated platform for researchers studying human lysine crotonylation.
• This work enhances the understanding of Kcr's role in physiology and pathology and facilitates further research.