Abstract: SA-PO0153
REN-Former Identifies Candidate Genes Associated with AKI-to-CKD Tubular-State Transition Supported by Human Genetics
Session Information
- Tubular, Metabolic, and Systemic Noncystic Genetic Kidney Diseases
October 24, 2026 | Location: Exhibit Hall A, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Genetic Diseases of the Kidneys
- 1202 Genetic Diseases of the Kidneys: Non-Cystic (Complex and Non-Cystic Monogenic)
Authors
- Mimura, Imari, Tokyo Daigaku Daigakuin Igakukei Kenkyuka Igakubu, Bunkyo, Tokyo, Japan
- Hosokawa, Sosuke, Tokyo Daigaku Daigakuin Igakukei Kenkyuka Igakubu, Bunkyo, Tokyo, Japan
- Hirakawa, Yosuke, Tokyo Daigaku Daigakuin Igakukei Kenkyuka Igakubu, Bunkyo, Tokyo, Japan
- Kawakami, Toshinaru, Tokyo Daigaku Daigakuin Igakukei Kenkyuka Igakubu, Bunkyo, Tokyo, Japan
- Ito, Masamichi, Tokyo Daigaku Daigakuin Igakukei Kenkyuka Igakubu, Bunkyo, Tokyo, Japan
- Kodera, Satoshi, Tokyo Daigaku Daigakuin Igakukei Kenkyuka Igakubu, Bunkyo, Tokyo, Japan
- Takeda, Norihiko, Tokyo Daigaku Daigakuin Igakukei Kenkyuka Igakubu, Bunkyo, Tokyo, Japan
- Nangaku, Masaomi, Tokyo Daigaku Daigakuin Igakukei Kenkyuka Igakubu, Bunkyo, Tokyo, Japan
Background
Acute kidney injury (AKI) is a major risk factor for subsequent chronic kidney disease (CKD), yet molecular regulators governing tubular-state transitions remain incompletely understood. Conventional single-cell RNA sequencing analyses identify differentially expressed genes between disease states but do not directly evaluate genes regulating directional state transitions. We developed a kidney-specialized foundation model, REN-former, to identify candidate regulators associated with transitions among Healthy, AKI, and CKD tubular states.
Methods
REN-former was generated by fine-tuning Geneformer using publicly available human kidney single-cell RNA-sequencing datasets from healthy controls, AKI, and CKD. In silico single-gene deletion and overexpression perturbation analyses were performed in proximal tubular cells to evaluate directional transcriptomic shifts among Healthy, AKI, and CKD-like states. Candidate genes were further integrated with kidney tubular eQTL and eGFR genome-wide association study datasets using summary-data-based Mendelian randomization (SMR) and colocalization analyses.
Results
REN-former identified distinct biological programs associated with tubular-state transitions. Loss-of-function perturbation identified programs related to stress resilience and adaptive repair, whereas gain-of-function perturbation identified inflammatory, oxidative stress, hypoxic, and proteostasis-related programs associated with transitions toward CKD-like states. Among 199 REN-former–prioritized genes tested by SMR analysis, six showed significant associations with kidney function after multiple-testing correction, and four genes (TTR, IFITM3, CALM1, and CALR) demonstrated colocalization between kidney tubular eQTL and eGFR-associated variants. CALR was identified in the Healthy-to-CKD overexpression program, whereas TTR, IFITM3, and CALM1 were identified in the CKD-to-Healthy overexpression program.
Conclusion
REN-former identified biologically distinct programs associated with tubular-state transitions and prioritized candidate genes supported by human kidney genetic data. These findings suggest that perturbation-enabled foundation modeling may provide a useful framework for identifying candidate regulators associated with CKD progression and tubular recovery.
Funding
- Private Foundation Support