Abstract
<title>Abstract</title> <p> <bold>Background</bold> : Psoriasis is characterized by widespread transcriptomic alterations, but traditional bioinformatic pipelines identify individual genes without evaluating multi-layered, network-level structures. We previously hypothesized that the coordinated activation and suppression of gene clusters on human chromosomes in psoriatic skin resemble a loss of genomic imprinting controls. Here, we present the first application of Structural Equation Modeling (SEM) in psoriasis genomics to evaluate this structural relationship. <bold>Methods</bold> : Genome-wide transcriptomic data from 180 skin biopsy profiles were obtained from the public NCBI GEO dataset GSE13355. Core gene families involved in epigenetic machinery and psoriasis pathology were curated using Gene Ontology and KEGG databases. Differential expression analysis was performed using the limma R package, selecting the top five most statistically significant genes as indicators per category. Expression profiles were Z-score normalized, and a two-latent SEM structure (Epigenetics and Psoriasis) was estimated via the lavaan R package. <bold>Results</bold> : The structural model demonstrated an excellent statistical fit (CFI = 0.973, TLI = 0.964, SRMR = 0.034, RMSEA = 0.087). A significant structural path coefficient of 1.00 (p < 0.001) connected the Epigenetics latent variable to the Psoriasis variable. The long non-coding RNA H19 exhibited a strong negative loading (-0.63), aligning with the suppression of the Wnt antagonist WIF1 (-0.88), the downregulation of CCL27 (-0.88), and the induction of the inflammatory driver CCL20 (0.93). <bold>Conclusion</bold> : Our structural findings support the hypothesis of altered imprinting-like controls in psoriatic tissue, closely linked to H19 downregulation. This study establishes SEM as a robust methodology for modeling complex, multi-latent genomic hypotheses. </p>