NullModels

The aim of a null models is to predict the expected effect of two drugs A and B given as the combined dose pair (a, b). The single response curves are here modeled as hill curves.

The following models are supported:

• Loewe
• Bliss
• Hand
• Tallarida (given as lower and upper bound due to non uniqueness)
• Highest Single Agent

Organization of the code:

• main.m: Run this script to read in data + evaluate data as well as to replicate the figures from our publication. You need to download the data from the supplementary of O'Neil et al. 2016 and change the values of singleDrugData =... (line 6) and CombinationData = ...(line 9) to the directories of the corresponding files on your coumputer.

Data structure

The data structure is given in the corresponding folder.

• Drug: All functionality corresponding Hill Curves (+ their fitting) and the single response data.
• Combination: Contains two drugs and gives all functionality corresponding to the null models and stores the combination treatment data.
• Cell Line: Contains Drug + Combination for one particular cell line (Hill Curves are fitted for each cell line separately).
• Data: Stores all data, that is read in (data comes from O'Neil et al. 2016),

Plots

The functions are in the folder Plots

• IsobolePlot: Isoboles of the different null models. Used in Figure 7. (Call IsobolePlot(D, 2, 3) ).
• CorrPlots.m: Correlates the predicion of different null models. Used in Figure 8. (Call CorrPlots(D.CellLines{3})) .
• VolumeMetricConceptPlot and VolPlot Plot volumes between the predicted and the measured response surfaces. (Not used in the publication. Call e.g. VolumeMetricConceptPlot and VolPlot(D)).

Publication

Sinzger, Vanhoefer, Loos, Hasenauer (2019) Comparison of null models for combination drug therapy reveals Hand model as biochemically most plausible. Scientific Reports, 9, 3002

If you have any questions regarding the implementations and/or need help: We encourage you to contact us!