Adapting a natural language processing tool to facilitate clinical trial curation for personalized cancer therapy

Area of Science:

• Bioinformatics
• Oncology
• Computational Biology

Background:

• Personalized cancer therapy relies on integrating complex molecular, pharmacological, and clinical data.
• Information overload and non-standardized data formats hinder oncologists' ability to access crucial patient-specific treatment information.

Purpose of the Study:

• To introduce informatics tools designed to streamline the retrieval and curation of cancer clinical trials focused on targeted therapies.
• To adapt and extend natural language processing (NLP) tools for improved annotation of clinical trial eligibility criteria.

Main Methods:

• Development and adaptation of an NLP tool for processing and analyzing cancer-related clinical trial information.
• Evaluation of the NLP tool using a curated dataset of 539 clinical trials as a gold standard.

Main Results:

• The informatics system demonstrated promising performance in facilitating the annotation of clinical trials.
• The system achieved 81% accuracy in predicting genotype-selected trials and an average recall of 0.85 for specific selection criteria.

Conclusions:

• The developed informatics tools show significant potential for enhancing the efficiency of personalized cancer treatment planning.
• The NLP-based approach offers good generalizability for identifying relevant clinical trials based on patient molecular profiles and selection criteria.