Fig. 8

Testing our inference method on simple networks. a For each network, numbered from 1 to 7, we simulated 100 cells using the full mechanistic model until the stationary regime was reached. Then we took a snapshot of their mRNA levels and inferred the parameters from this data. The result was called successful when the inferred structure (topology and nature of the links) was the same as the true network. b For each network (rows), 10 datasets were simulated and the results were reported by counting the number of inferred θ corresponding to each structure (columns), highlighting successes (blue) and failures (orange). The perfect inference would lead to 10 for all the diagonal terms and 0 everywhere else. c Examples of simulated mRNA datasets (one for each network). Although having coherent signs, Pearson’s correlation coefficients (top right of each plot) would clearly be insufficient to distinguish between the different networks