CƠ SỞ DỮ LIỆU GEN VÀ GIẢI TRÌNH TỰ THẾ HỆ MỚI PHÁT HIỆN ĐỘT BIẾN ĐÁP ỨNG THUỐC UNG THƯ PHỔI TẠI VIỆT NAM

Abstract

Objective: To establish a lung cancer genetic database tailored for Vietnamese patients by applying artificial intelligence (AI) and next-generation sequencing (NGS) for the detection of therapy-related variants. Methods: A cross-sectional descriptive study was conducted using PubMed literature data and samples from 21 patients with advanced non-small cell lung cancer (2024–2025). NGS was performed on tumor tissue samples, while AI and natural language processing (NLP) were employed to filter and extract information from international databases (ClinVar, COSMIC, OncoKB, etc.). Results: The system screened over 400,000 publications, extracted more than 9 million entities, and identified 97 genes associated with pathogenesis, prognosis, and therapeutic response, including EGFR, KRAS, ALK, BRAF, and TP53. In 21 sequenced patients, 663 variants were detected; 12 variants in 11 patients were classified as pathogenic or likely pathogenic with therapeutic relevance. The most frequent were EGFR del19 (66.7%) and EGFR L858R (16.7%). Additionally, one case carried concurrent del19 and T790M variants, associated with TKI resistance. Variant allele frequencies (VAF) ranged from 5-66%, reflecting tumor heterogeneity. Conclusion: The integration of AI and NGS proved effective in mining genetic data for lung cancer and provided an initial foundation for a population-specific database in Vietnamese patients.