Proximity-Aware Transcriptomics Reveals Non-Random Genomic Clustering of Low-Annotated ncRNAs Near Protein-Coding Genes in SMA iPSCs

To overcome this limitation and systematically decode this uncharacterized transcriptome, we developed the Proximity-aware ENSG Pipeline (PEP), a multi-layered computational framework that integrates differential expression processing, Ensembl-based biotype classification, linear genomic proximity modeling, and chromosome-conditional permutation testing (10,000 iterations).We deployed PEP on total RNA-seq data generated from a stage-matched cohort of human induced pluripotent stem cells (iPSCs) derived from patients with Spinal Muscular Atrophy (SMA) and healthy controls.

PEP successfully captured the primary causal genetic lesion of SMA (SMN1 down-regulation; log2FC = -3.22, P < 0.001) validating the biological baseline. Our framework demonstrated that low-annotated ncRNAs (n = 9.070) form a highly organized, gene-centered class-level architecture, remaining non-randomly proximal to protein-coding genes with an observed median neighbor distance of 29.3 kb versus a random null expectation of 105 kb (P < 0.001).