Epigenetic target identification strategy based on multi-feature learning.

Area of Science:

• Biochemistry
• Computational Biology
• Pharmacology

Background:

• Epigenetic drugs are increasingly vital in cancer treatment.
• Identifying epigenetic targets for bioactive compounds is crucial for drug discovery.
• Advances in chemogenomic data necessitate improved target identification methods.

Purpose of the Study:

• To introduce a novel epigenetic target identification strategy (ETI-Strategy).
• To leverage machine learning for predicting protein-ligand interactions and identifying epigenetic targets.
• To enhance the accuracy and efficiency of epigenetic drug discovery.

Main Methods:

• Integration of a multi-task graph convolutional neural network prior model.
• Utilization of a protein-ligand interaction classification discriminating model.
• Training and validation using large-scale bioactivity data for 55 epigenetic targets.

Main Results:

• Achieved an AUC of 0.934 for the prior model and 0.830 for the discriminating model.
• Outperformed inverse docking in predicting protein-ligand interactions.
• Demonstrated superior accuracy compared to other open-source target identification tools.

Conclusions:

• The ETI-Strategy effectively utilizes molecular and protein-level information for accurate activity prediction.
• Machine learning significantly contributes to identifying potential epigenetic targets.
• This approach offers a novel pathway for epigenetic drug discovery and development.