Non-linear combinatorial analysis of blood transcriptomes identifies PRKAR1A as a regulator of TDP-43 pathology in amyotrophic lateral sclerosis

Abstract Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by progressive loss of motor neurons. Accurate and accessible blood-based diagnostics for neurodegenerative diseases, including ALS, are being progressively required. Although blood cell gene expression profiles have potential clinical utility for distinguishing ALS, robust transcriptomic biomarkers for supportive diagnosis have not yet been established. Here, we analyzed publicly available peripheral blood mononuclear cell (PBMC) transcriptomic data from ALS patients using Maximum Mean Discrepancy (MMD), a kernel-based method that captures non-linear distributional differences in a reproducing kernel Hilbert space and enables the extraction of informative gene combinations while minimizing multicollinearity, a common issue in multiple regression models. Using this approach, we identified a non-linear three-gene combination—PRKAR1A, QPCT, and TMEM71—that distinguished ALS from healthy controls with an area under the curve (AUC) of 0.83 in a public PBMC dataset. This achievement was confirmed in laboratory PBMC samples with an AUC of 0.85, supporting the robustness of the identified gene signature in independent samples. Furthermore, these genes also enabled ALS classification in induced pluripotent stem cell (iPSC)-derived motor neurons with an AUC of 0.79. Knockdown of PRKAR1A, QPCT, or TMEM71 in motor neurons increased the TDP-43 expression levels, and PRKAR1A knockdown induced the mislocalization of TDP-43 accompanied by phosphorylation, suggesting a potential link to ALS-related pathophysiology. These findings suggest that non-linear gene combinations may provide a useful strategy for identifying blood-based biomarkers and offer insights into ALS pathogenesis. This non-linear, data-driven analytical framework enabled the transition from unbiased gene discovery to the identification of pathophysiology-associated molecules by in vitro functional validation.