Bridging data gaps: Methodological advances in extracting and analyzing genetic information from unstructured clinical records in hereditary cancer

Abstract

Introduction

The global impact of cancer, driven by both acquired and hereditary mutations, underscores the necessity for extensive research efforts. Despite the increasing volume of genetic data, significant gaps remain in data science research, particularly in Latinos and admixed populations. This study utilizes advanced data science techniques to integrate genetic and clinical data, aiming to improve the understanding of hereditary cancer in Colombia and demonstrating the transformative potential of data-driven approaches in cancer research.

Methods

This observational study analyzed healthcare databases from four regions and 11 cities in Colombia. Genetic data were extracted from PDF reports within SURA Colombia's Electronic Health Records (a Latin American health insurance provider) for individuals referred for hereditary cancer testing between October 2019 and November 2021. Variants in 30 genes, aligned with NCCN guidelines, were examined using Next-Generation Sequencing (NGS). Data extraction was automated using Python and R, followed by integration and analysis of genetic, clinical, and sociodemographic data using advanced data science tools hosted on Azure infrastructure. These tools enabled predictive modeling and cross-referencing to explore correlations between genetic variants and clinical outcomes.

Results

The study included 1377 patients, with a predominance of women (92.81 %) and 63 % from the northwestern region of Colombia. The largest age group (40.37 %) was between 31 and 44 years, and 95.35 % had a personal cancer history, primarily breast cancer (75.86 %). Hereditary cancer testing revealed 145 positive results and 587 uncertain outcomes. Data science-driven analysis identified higher positivity rates in patients aged 31–44 and over 50, particularly in the northeast and central regions. Among positive results, 42.6 % included variants of uncertain significance, with 95.9 % of these patients having a personal cancer history.

Conclusion

This study highlights the significant role of data science in analyzing hereditary cancer data. Advanced computational techniques can aid in genetic variant reclassification, uncover patterns in underrepresented populations, and inform personalized interventions for hereditary cancer management in Latin America.