Which genes actually steer a T cell?
A large effect can still be the wrong answer
Where did the T cell go?
Direction and repeatability add signal beyond magnitude
The signal is real and biologically narrow
The property is immune-contextual, not universal
Controller is not target or biomarker
The product knows when to stop
A prospective experiment can falsify the next claim
Executable, reusable, auditable
01 / 10 Which genes actually steer a T cell?
00:00
T-CTRL separates perturbation magnitude from functional direction and cross-donor repeatability. The authoritative full-sample gain is +0.357 AUPRC. The separate, leakage-free out-of-fold gain is +0.215. The broad external regulator test is FAIL at −0.281; CAR-T response prediction is NULL at AUROC 0.533, p=0.138; missing scGPT inputs are NOT-EVALUABLE. Controller does not mean therapeutic target or validated biomarker.