The penalized generalized linear models are generally very good, and provide the added advantage of easy interpretation and relatively low model complexity; as noted in the previous paragraph, a softer regularization might be beneficial in the future.

As a linear transformation, the standardization does not affect linear models , though the additional preprocessing truncation to 60 has an appropriate impact on outliers.

To assess how much this discrimination depends on the classification approach utilized rather than the underlying information content in the data, we employed three different representative classification techniques: penalized logistic regression (a regularized generalized linear model based on Lasso), regularized random forest (a tree-based model), and support vector machine (a kernel-based model).

As With classification, the linear model dominates, and all methods perform similarly well With any of the input feature sets.

Once again the linear model dominates the nonlinear models, particularly for ADCC.

The linear model is parameterised by the expression at time 0 (p 1) and the change in expression (p 2) from which the rate of increase or decrease can be calculated.

The inference of model parameters from CAGE data for the early peak and linear models using nested sampling and the 11 based likelihood is illustrated in Fig 1C.

CAGE clusters are assigned to one of the exponential kinetic signatures if log Z for that signature is greater than 10 times log Z for the linear model and log Z minus its standard deviation (sd) is greater than log Z plus the estimated sd for any other eXponential signature (nested sampling computes log Z for parameters mapped to O..1 and we used the resulting log Z for the unit cube for model comparison).

CAGE clusters were assigned to one of the exponential kinetic signatures or to the linear model according to the value of log Z.

An example of fitting early peak and linear models to an EGR1 time course is presented in Fig 1C.