Graph regularized non-negative matrix factorization with norm regularization terms for drug-target interactions...

Area of Science:

• Bioinformatics
• Computational Drug Discovery
• Cheminformatics

Background:

• Identifying drug-target interactions (DTIs) is crucial for drug development.
• Wet experiments for DTI identification are resource-intensive.
• Computational methods accelerate drug discovery but often neglect feature sparsity and algorithm convergence.

Purpose of the Study:

• To propose an accurate computational method for predicting DTIs.
• To address limitations in existing non-negative matrix factorization methods for DTI prediction.

Main Methods:

• Developed iPALM-DLMF, a novel method for DTI prediction using non-negative matrix factorization.
• Incorporated graph dual regularization for drug and target similarity integration.
• Utilized L1 norm regularization for feature matrix sparsity.
• Employed an inertial Proximal Alternating Linearized Minimization for model convergence.

Main Results:

• iPALM-DLMF demonstrated superior performance compared to state-of-the-art methods.
• Case studies showed high validation rates: 46/50 for gabapentin targets and 47/50 for prostaglandin-endoperoxide synthase 2 targets.
• The method effectively integrates similarity information and ensures feature sparsity.

Conclusions:

• iPALM-DLMF offers a more accurate and efficient approach to predicting DTIs.
• The method's convergence properties and regularization techniques enhance predictive power.
• This computational tool can significantly aid in accelerating drug discovery processes.