Real-world evidence on data quality in precision oncology platforms: insights from the Molecular Twin Research...

The increase in precision oncology largely relies on the availability of high-quality, longitudinal, and multi-source, multi-purpose, and multidimensional corresponding data that integrate clinical, pathological, and molecular information. While new and advanced biomedical methods enable large-scale data generation, the operational challenges associated with data retrieval, harmonization, and quality control remain insufficiently described in scientific literature. In parallel, data incompleteness and heterogeneity in collection practices and coding standards are diminishing confidence in precision medicine programs. This methodological case study focused on the challenges of data harmonization within the Molecular Twin Research Umbrella Protocol at Cedars-Sinai Medical Center. This systematic four-stage process includes requirements for data access and assembly, processing for cleaning and harmonization, and quality verification before distribution. Pairwise comparisons have been made for each type of record, from electronic health records through cancer registration and biobanking. The demographics showed a very high concordance (>95%), whereas the clinically essential variables, such as tumor TNM stage, diagnostic specificity, and intervention schedules, showed moderate discordance (14.8%-17%). These discrepancies are a significant hindrance to the readiness of the cohorts for prediction models. Based on these results, we propose mitigation strategies aimed at improving the accuracy, completeness, and standardization of longitudinal oncology cohort datasets. This study provides key recommendations for long-term oncology cohorts and for the development of digital twin infrastructure for any institution, stressing that sound data quality infrastructure is a cornerstone of trustworthy precision oncology and translational research.