Fig. 4

Study of the virtual chondrocyte state transition and in silico screening of target perturbations. A Relation between Inflammation and TGFβ and influence on the chondrocyte state. Perturbations are applied on the healthy attractor, bar height gives the average percentage of transitions towards one of the target states, error bar denotes standard deviation. “Infl.” refers to imposed inflammation, “TGF” refers to TGFβ over-activation, and “Alk balance” to the modification of the ratio between TGFβ receptors (ALK1 and ALK5). Conditions were mimicked as described in the table. “−” denotes no modification of the initial value. A transition from “healthy” to “healthy” means no transition. B All single node perturbations triggering a state transition from the Healthy (resp. Hypertrophic) state. (C) Markov chain providing the overall probability of transition from one state to another, under single node perturbations. Arrows indicate transitions from an initial state towards a target state with the associated probability. Thus, the total probability of outgoing arrows for any state is 1.0. D PCA visualizing the results of the systematic screening of all possible combinatorial perturbations on a hypertrophic-like chondrocyte. Each dot represents one of the 7080 screened conditions. Principal components are computed based on the percentage of transitions towards the 3 attractors, reported as eigenvalues (blue arrows). Dot colors correspond to threshold in the percentage of transitions towards the healthy state for potential OA therapies. The details of the predicted conditions leading to 70% and up to 100% of transitions towards the healthy state are available in Additional file 9, Data S4, the ones that were selected for experimental validation are further described below.