Figure 1

The CellNOptR framework. A. A CellNOptR analysis takes as input 2 text files: (1) a Prior Knowledge Network (PKN) as a SIF file[39], (2) a dataset in the MIDAS format ([34], see Figure4). The package then maps the data onto the PKN, processes the network and trains the resulting model. CellNOptR outputs a series of HTML pages containing the summary of the analysis, hyperlinked to diagnostic graphs, and the trained networks. Multiple logic formalisms can be used for the training. The CellNOptR package implements most of the workflow and the simplest Boolean logic steady-state (1 or 2) approach. B. Only steps that are specific to a particular logic formalism are coded in add-on packages. CNORfuzzy implements a constrained fuzzy logic steady-state approach[35]. CNORdt fits time course data using a Boolean representation of the states of nodes and a synchronous update simulation scheme. CNORode fits detailed time course data by deriving and training continuous logic-based ordinary differential equations. C. The choice of a logic formalism depends on the data at hand and the modeling goals: with no time course data, the user can choose between the two steady-state implementations (CNORfuzzy and CellNOptR) based on the size of the network, richness of data and suspected impact of partial effects. If very limited time course data is available, users can use the Boolean 2 steady-states implementation in CellNOptR. With detailed time course data, one can choose between the Boolean discrete time implementation in CNORdt and the continuous ODE based implementation in CNORode, mainly based on the complexity of the network and the richness of the data. For the networks, the following color conventions are used: for nodes: green=stimulated, red=inhibited, blue=measured, dashed=compressed; edges (referring to the optimised model): green=present at time 1, blue=present at time 2, grey=absent, dashed edge=compressed.