LSD-pipeline: Causal Inference of miRNA Network Effects in Alzheimer's Disease

MicroRNAs (miRNAs) are linked to Alzheimer's disease (AD), yet prior studies have examined individual miRNAs and their direct targets. Current methods for understanding miRNA regulation in AD reveal correlations, not causal links, and few techniques estimate multi-stage networks from miRNAs to target gene transcription factors (TFs).

Researchers have created the LSD pipeline (LASSO-SEM-DoWhy) to identify and validate miRNA causal pathways in AD by combining LASSO feature selection, multi-stage structural equation modeling, and DoWhy causal inference. Using six blood miRNA and brain mRNA datasets, the LSD pipeline identified four validated miRNAs (miR-30d-5p, miR-92a-3p, miR-296-5p, miR-193a-5p) as potential AD biomarkers, achieving 86% ROC accuracy in an independent cohort.

Several miRNAs with no direct association with AD showed significant effects through their target networks, while direct analyses were not supported at the network level, highlighting the importance of network-level analysis. By extending to the TF layer, complete miRNA-target TF cascades were identified, with HMGA1, NKX2-3, and PRRX2 acting as key intermediaries.

Classic pathways confirmed the primary roles of miRNAs in tau pathology and synaptic dysfunction. The LSD pipeline's findings suggest that miRNA effects are largely age-independent, indicating that miRNAs may initiate AD pathogenesis early on. Although primarily focused on AD, the LSD pipeline offers a generalizable approach for uncovering causal regulatory mechanisms in other diseases.