Genetic Ancestry and Cancer Genomics: An LLM-Based Framework for Equitable Precision Oncology

The incidence of cancer, its genetic mutations, and treatment effectiveness depend heavily on genetic ancestry. Yet existing cancer genomics datasets underrepresent non-European populations, which restricts the application of precision oncology results to diverse patient groups. In this study, a retrieval-augmented generation (RAG) framework used large language models (LLMs) to analyze 30 PubMed-indexed cancer genomics studies that contained ancestry-related underrepresented groups in genomic research. The system used document chunking, embeddings, and semantic retrieval to find connections between ancestry, mutations, and treatment effects. Our system achieved 92% retrieval precision and 0.87 F1-score in mutation-ancestry mapping, validated against The Cancer Genome Atlas (TCGA) datasets. The platform successfully detected how different ancestry groups relate to cancer mutations and drug responses through its analysis of TP53 mutations in African breast cancer patients and better EGFR-inhibitor effectiveness in East Asian lung cancer patients. The study revealed two important findings, which showed that people of African ancestry have a higher risk of clear-cell renal carcinoma when they carry VHL variants, and survival rates differ based on ancestry. The research proves that AI-based literature analysis can merge scattered ancestry-cancer data to discover fresh connections between genetic background and cancer. The implementation of ancestry-based frameworks in precision oncology practice will help achieve fair cancer treatment for all patients.