Homogeneous network inference for subtype-stratified master regulator analysis in cancer

Fixed-size Spearman regulon networks inferred for 34 TCGA cancer-by-subtype cohorts identify master regulators, but subtypes with fewer than 60 tumours do not reliably reproduce them.

• Why it matters: Master-regulator analysis needs one regulatory network per subtype cohort, yet mutual-information networks are unavailable for the smallest cohorts, risking inconsistent methods across subtypes.
• What they did: They infer every cohort’s network using a single fixed-size Spearman regulon pipeline: Spearman rank correlation with rho ≥ 0.2, corto DPI pruning, bootstrap support, and truncation to the 100 best-supported targets per regulator.
• The result: Rank correlation is more powerful than simulated MI estimators at n = 95, overlapping in 0.7 of top-20 regulators, but networks from fewer than 60 tumours fail to reproduce their own cohort regulators, and cancer-wide networks do not rescue small subtypes.

AI summary of the abstract · may contain errors · not medical advice · read the paper.