NfκBin: a machine learning based method for screening TNF-α induced NF-κB inhibitors.

Area of Science:

• Computational chemistry and drug discovery
• Bioinformatics and cheminformatics
• Molecular biology and immunology

Background:

• Nuclear Factor kappa B (NF-κB) signaling is implicated in chronic inflammatory diseases like rheumatoid arthritis, inflammatory bowel disease, and asthma.
• Targeting NF-κB is a key strategy for developing novel therapeutics for these conditions.
• Developing predictive models for NF-κB inhibitors can accelerate drug discovery.

Purpose of the Study:

• To develop and validate a computational approach for predicting drugs that inhibit TNF-α induced NF-κB signaling pathways.
• To identify potential NF-κB inhibitors from a library of FDA-approved drugs using a machine learning model.
• To provide a reliable tool for researchers investigating NF-κB related diseases.

Main Methods:

• Utilized a dataset of 1,149 inhibitors and 1,332 non-inhibitors from PubChem.
• Computed chemical descriptors using PaDEL software and selected relevant features via univariate analysis and SVC-L1 regularization.
• Constructed and optimized machine learning models, including a support vector classifier, achieving a maximum AUC of 0.75.

Main Results:

• Machine learning models using 2D and 3D descriptors, and molecular fingerprints showed varying performance (AUC 0.56-0.66).
• Feature selection significantly improved model performance, with the support vector classifier achieving an AUC of 0.75.
• The best-performing model successfully screened FDA-approved drugs, with predicted inhibitors largely aligning with known experimental findings.

Conclusions:

• The developed computational model demonstrates high predictive reliability for identifying NF-κB inhibitors.
• This approach can accelerate the discovery of novel therapeutic agents for chronic inflammatory diseases.
• The models are available via a standalone software and web server (NfκBin) for broader accessibility.