Underlying mathematical model
Linear models, widely spread in different fields of science, provide a simple and easily understandable design but over-simplify the underlying biological processes. Nonlinear behavior, e.g., saturation effects, cannot be represented. Considering that the number of available samples is typically relatively small, the incorporation of many different data types and according parameters into the model could result in excessively complex designs prone to overfitting, but this issue lacks general awareness. Only two of the presented methods incorporate parameter priors (ISMARA and biRte), and two apply cross validation techniques to estimate prediction performance (method by Schacht et al. and RACER). Further, the effect of temporal buffering between TF binding and the actual effect on gene expression is not included in any of the methods.