Natural Language Processing-Assisted Literature Retrieval and Analysis for Combination Therapy in Cancer.

Area of Science:

• Biomedical Informatics
• Computational Biology
• Drug Discovery

Background:

• Anticancer drug development faces challenges in achieving durable responses.
• Rational drug combinations offer synergistic effects and can overcome drug resistance.
• Efficient knowledge discovery in combination therapy is crucial for advancing cancer treatment.

Purpose of the Study:

• To develop RetriLite, an information retrieval and extraction framework for identifying relevant research on anticancer combination therapies.
• To facilitate knowledge discovery and reduce manual review of biomedical literature.
• To create a publicly accessible search engine for combination therapy data.

Main Methods:

• Developed RetriLite, a framework using natural language processing and a domain-specific knowledgebase.
• Customized RetriLite into an informatics pipeline to identify efficacy of combination therapies in clinical and preclinical studies.
• Leveraged modular architecture for synergistic information retrieval and NLP techniques.

Main Results:

• RetriLite achieved a high F-score of 0.93 in a pilot study.
• Demonstrated high accuracy in identifying relevant papers (95.9% true positive) and distinguishing clinical from preclinical studies (97.6% accuracy).
• Publicly accessible online search engine with an intuitive interface was launched.

Conclusions:

• RetriLite provides a flexible framework for domain-specific information retrieval and extraction systems.
• The system enhances knowledge discovery in combination therapy through metadata tags and keyword highlighting.
• Facilitates more effective and efficient research for scientists in the field.