AI-driven drug discovery using a context-aware hybrid model to optimize drug-target interactions

Area of Science:

• Computational chemistry and cheminformatics
• Artificial intelligence in drug discovery
• Bioinformatics and computational biology

Background:

• Drug discovery is a complex, costly, and time-consuming process with high failure rates.
• Identifying suitable drug candidates and predicting drug-target interactions remain significant challenges.
• Existing predictive models often lack the necessary accuracy and adaptability for efficient drug discovery.

Purpose of the Study:

• To propose a novel Context-Aware Hybrid Ant Colony Optimized Logistic Forest (CA-HACO-LF) model for enhanced drug-target interaction prediction.
• To improve the accuracy and efficiency of candidate selection in the pharmaceutical drug discovery pipeline.
• To leverage AI and machine learning techniques for optimizing drug discovery processes.

Main Methods:

• Utilized a Kaggle dataset of over 11,000 drug details.
• Applied text normalization, stop word removal, tokenization, and lemmatization for data pre-processing.
• Employed N-grams and Cosine Similarity for feature extraction and semantic relevance assessment.
• Integrated Ant Colony Optimization (ACO) with Random Forest (RF) and Logistic Regression (LR) for classification.

Main Results:

• The CA-HACO-LF model achieved superior performance in predicting drug-target interactions.
• Demonstrated high accuracy (0.986%), precision, recall, F1 Score, and Cohen's Kappa.
• Outperformed existing methods across multiple evaluation metrics, including RMSE, AUC-ROC, MSE, and MAE.

Conclusions:

• The CA-HACO-LF model offers a significant advancement in AI-driven drug discovery.
• Context-aware learning enhances the model's adaptability and predictive power.
• This approach promises to accelerate the identification of effective drug candidates and optimize drug development pipelines.