A Systematic Review of Drug-Related InteractionsUtilizing Deep Learning and LLMs for Prediction and Mitigation.

Area of Science:

• Computational drug discovery
• Pharmacogenomics
• Bioinformatics

Background:

• Computational drug discovery accelerates treatment screening and reduces costs.
• Machine learning (ML), deep learning (DL), and large language models (LLMs) are transforming the field.
• Key areas include drug-drug interactions (DDIs), drug-target interactions (DTIs), and adverse drug reactions (ADRs).

Purpose of the Study:

• To systematically review ML and DL techniques in drug discovery.
• To focus on the extraction of DDIs, DTIs, and ADRs.
• To identify research gaps and propose future directions.

Main Methods:

• Systematic literature review of over 100 papers (2020-2025).
• Categorization of methods into deep learning, machine learning, graph learning, and hybrid models.
• Analysis of natural language processing (NLP) and LLMs for data extraction.

Main Results:

• ML and DL significantly impact DDI, DTI, and ADR prediction.
• NLP and LLMs show transformative potential in extracting insights from biomedical and chemical data.
• Key databases and datasets for drug discovery were identified.

Conclusions:

• Existing research often lacks simultaneous DDI, DTI, and ADR extraction.
• A more holistic approach is needed to address these interconnected aspects.
• Performance metrics of various models were evaluated, highlighting their strengths and weaknesses.