Identifying Associations between Somatic Mutations and Clinicopathologic Findings in Lung Cancer Pathology Reports

Area of Science:

• Bioinformatics
• Computational Biology
• Oncology

Background:

• Non-small cell lung cancer (NSCLC) treatment is increasingly guided by genetic mutation profiles.
• Pathology reports contain rich clinical information often in unstructured text.
• Integrating clinical findings with genetic data is crucial for personalized medicine.

Purpose of the Study:

• To develop an informatics methodology for identifying statistically significant associations between clinical findings in NSCLC pathology reports and actionable genetic mutations.
• To create a pipeline leveraging NLP and semantic analysis for this integration.

Main Methods:

• Developed an information extraction and analysis pipeline using NLP, biomedical terminologies, semantic similarity, and clustering.
• Applied the pipeline to 142 NSCLC lobectomy surgical pathology reports with NGS data.
• Focused on mutations in EGFR, KRAS, BRAF, and PIK3CA oncogenes.

Main Results:

• Identified 732 distinct positive clinical concepts from unstructured pathology reports.
• Highlighted multiple statistically significant associations (P ≤ 0.05) between clinical findings and specific gene mutations.
• Validated findings against existing published literature, confirming consistency.

Conclusions:

• An automated pipeline was developed to associate clinical findings from unstructured text with genetic mutations.
• The methodology is generalizable to other cancer types and clinical reports.
• This approach can advance the understanding of genetic mutations in cancer development and treatment.