Fig. 7: Overview of the developed approach.

At the core of the method is our connected model. Outside the core connected model, there exists phosphoproteomic data (gray circles) not covered by the model. Using lists of interactions and our automatic model expansion algorithm we are able to add parts of this phosphoproteomic data, for proteins adjacent to the core model to create an expanded model that can be used to simulate scenarios. In step 1, the method find all phosphosites that are adjacent to the model using a set of allowed interactions. The possible additions are subdivided into independent parallel subproblems. In step 2, the methods tests the possible additions, first using only a single input in step 2a, then using dual inputs in step 2b. The additions are evaluated based on the agreement to data, exemplified using the two time-series plots, where dots with error bars correspond to experimental data and solid lines correspond to simulations of the new addition using inputs from the model. If the agreement is good enough, the potential addition is added to the model. After step 2, all approved additions are collected and added as a new layer in the model in step 3. Including the new layer, the method checks for additional possible additions. If any adjacent site exists, the method returns to step 1 and finds all possible additions. If no possible additions exist, the method allows additional interactions of a lower confidence in step 4 and again checks for possible additions. When interactions of all confidence levels have been included and there exists no more adjacent sites that can be added to the model, the expansion is stopped. Finally, the expanded model can used to simulate new scenarios in step 5.