Gene Network Inference Lab

A hidden ground-truth gene regulatory network of ten genes drives a simulated Hill-function expression time series, exactly like real RNA-seq or scRNA-seq data — but you never see the wiring, only the noisy readings. This lab lets you reconstruct that network with three real inference strategies used in genomics — correlation thresholding, partial-correlation regression, and CRISPR-style single-gene perturbation testing — then scores every inferred edge against the true network in 3D: green for correctly recovered edges, red for confounder-driven false positives, gray-dashed for edges the method missed. Tuning measurement noise, time-series length, and a toggleable hidden confounder shows exactly why perturbation data is considered the gold standard for validating GRNs, and why correlation alone is not: precision, recall and F1 update live as you switch methods.