Let us con sider that a drug i with target set T0 and EC50 profile ei,1, ei,2, ei,n is applied at concentration x nM. For each EC50 value ei,j, we can fit a hill curve or a logistic func tion to estimate the inhibition of target j at concentration x nM. For instance a logistic function will estimate the drug target profiles Imatinib Mesylate for a combination of drugs at differ ent concentrations. To arrive at the sensitivity prediction for a new target inhibition profile, we can apply rules sim ilar to Rules 1, 2 and 3 along with searching for closest target inhibition profiles among the training data set. The block analysis performed using discretized target inhi bitions can provide smaller sub networks to search for among the target inhibition profiles.

Incorporating network dynamics in the TIM formulation The TIM developed in the previous sections is able to predict the steady state behavior of target inhibitor com binations but cannot provide us with the dynamics of the model or the directionality of the tumor pathways. This limitation is a result of the experimental drug perturbation data being from the steady state. Our results show that the proposed approach is highly successful in locating the primary faults in a tumor circuit and predict the possible sensitivity of target combinations at the current time point. However, exten sion of this model to incorporate the directional pathways will require protein or gene expression measurements. The extension refers to steps F1 and F2 in Figure 1. These steps are not necessary to design the control policy but if performed can provide superior performance guarantees.

If we plan to infer a dynamic model from no prior knowl edge, the number of required experiments will be huge and will primarily require time series gene or protein expression measurements. In this section, we will show that the circuit produced by our TIM approach can be used to significantly reduce the search space of directional pathways. To arrive at the potential dynamical models sat isfying the inferred TIM, we will consider the possible directional pathways that can generate the inferred TIM and convert the directional pathways to discrete Boolean Network models. The TIM can be used to locate the feasible mutation patterns and constrain the search space of the dynamic models generating the TIM.

For the duration of the Network Dynamics analysis, we will consider the two dynamic models shown in Figure 4. Dongri Meng Dongri Meng inhibition of target j as f 1 Note that at concentration x ei,j, f 0. 5 as desired. This approach can Entinostat be applied to arrive at a continuous target profile zi,1, zi,2, zi,n of a drug that is dependent on the applied drug concentration. The zi,js denote real numbers between 0 and 1 representing the inhibition ratio of target j.