Back-Translation: Novel Approach to Target Identification for CKD via Integration of Clinical and Biobank Data

Background: Despite innovative treatments becoming available for Chronic Kidney Disease (CKD), there remains a high residual risk of CKD progression indicating the need for novel therapeutic options. Recently, the increased availability of population-scale genomic data has provided a boost for data-driven approaches to drug discovery. However, biobanks often lack comprehensive longitudinal phenotyping and crucial clinical information. In contrast, clinical trial data are often sparse on omics information, hindering mechanistic understanding of clinical findings. In this project, we leverage both data sources by “back-translating” findings on biomarker signatures of CKD fast progression from clinical trial data into biobank data for generation of novel therapeutic hypothesis. Methods: First, we developed a Cox regression model to describe a subgroup of CKD fast progressors within the clinical trial data. Resulting model coefficients were then used to construct a risk score for fast progression which can be translated to biobanks, thereby enabling the conduct of variant-based and gene-based association tests. Results: We demonstrate the value of the back-translation approach using clinical trial data from the FIDELITY dataset (N=12,990) combined with biobank data from the UK Biobank (N=502,148) and, for validation purposes, the German Chronic Kidney Disease Study (N=5,217). Our results show that the derived risk score, which includes established risk markers like hemoglobin or UACR, accurately identifies high-risk patients in both FIDELITY (Concordance index: 0.8089) and GCKD (Concordance index: 0.802). Our genetic analysis in UK Biobank identifies 27 genetic regions from the variant-based analysis and 4 hits from the gene-based analysis which may serve as the foundation for a novel therapeutic target investigation for CKD fast progression. Conclusion: We propose an approach to target identification that offers an innovative opportunity to integrate clinical data with population genetics data for target identification via "back-translation," utilizing clinical insights previously underutilized in a research context. By bridging clinical and genetic data, our framework enhances the potential for discovering novel therapeutic targets and for advancing precision medicine. Funding: Commercial Support - Bayer Pharma