Can TargetDiscovery be used to estimate non-monotone dose response curves?
No. TargetDiscovery aims to estimate a linear effect of the target variable (e.g., dose of a drug) on the outcome. In other words, the reported impact of changing the target variable from level “a” to level “b” on the mean outcome does not depend on the initial level “a” but only on the difference “b-a”. If the true dose response curve is not linear, the reported estimate can be interpreted as an approximation to the true curve. This approximation is reasonable if the true dose response curve is monotone. If it is not, the results may be misleading.