Abstract / Summary
Chronic Kidney Disease (CKD) progresses through a variety of molecular mechanisms that are not fully captured by routine measures of tissue injury. Recently, viral mimicry the activation of endogenous retroviral elements and innate immune sensors has emerged as a key driver of sterile inflammation and fibrosis. Despite the presence of various molecular factors driving the progression of CKD, the challenge in therapeutic targeting has been the inability of conventional clustering to reflect the underlying pathomechanisms or to translate epithelial dedifferentiation into therapeutic targets. Therefore, there is a requirement for reliable Integrated Artificial Intelligence (AI)-Driven Virtual Topological Mapping Framework that can decipher the underlying disease niches from high-dimensional transcriptomic datasets. This research analysed approx 20,000 individual nuclei transcriptomes (snRNA-seq) of patients accessible on GEO database with progressive IgA Nephropathy with a multi-omics computational model. Trajectories were then traced using Partition-based Graph Abstraction (PAGA) and Diffusion Pseudotime (DPT). The Random Forest Classifier (n=1000) yielded transcriptomic drivers, which were then navigated on a 100*100 virtual spatial units spatial grid. Using this pipeline, we validated our predictive findings using a patient-level hold-out testing partition in a rigorous way by utilizing statistical enrichment methods (P adj < 10 -7 ) to ensure accurate results are produced. Connectivity mapping was then used for virtual exploration for drug repositioning on the candidate hubs. Our results showed Trajectory inference revealed a synchronous ''Pseudo-Viral'' program, which is initiated by innate immune sensors in dedifferentiating epithelial cells. Our machine learning framework achieved a diagnostic Area Under the Curve (AUC) of 0.914, with a single ''SARS-CoV-2 Mimicry Hub'' emerging as the top driver of renal decline. Spatial projection pinpointed these signals to a discrete Pathogenic Epithelial Cluster at (X:50, Y:50), where the density of the Pseudo-Viral Niche reached a normalized peak of 1.0. Drug repurposing identified Baricitinib as a computationally predicted therapeutic candidate with a >0.95 Transcriptomic Reversal Score, requiring future experimental validation. Our in silico model hypothesizes that the progression of IgA-nephropathy-associated CKD is strongly associated with a 'viral mimicry' transcriptomic signature occuring in a topologically conserved virtual niche called the Pathogenic Epithelial Cluster. The Pathogenic Epithelial Cluster now serves as a target for next-generation aptamers and small molecules.