The PAR(2) Discovery Engine applies AR(2) autoregressive modelling to gene expression time series data. The core metric is the eigenvalue modulus |λ|, which measures temporal persistence — how self-sustaining a gene's expression dynamics are over time. Core clock genes (BMAL1, CLOCK, PER1/2, CRY1/2, NR1D1) show consistently higher |λ| than clock-controlled target genes across 13 tissues, 4 species, and 22 independent datasets. The clock–target gap is pre-specified and replicates in mouse, human, baboon, and yeast. The platform includes a genome-wide coupling scan across ~21,000 genes, an ODE model zoo for theoretical validation, disease comparisons (Alzheimer's, colorectal cancer, p53 regulon), and a Discovery Engine for uploading custom gene expression CSV files.