tification of the clinically relevant closest common ancestors between the HHM and the AHM. Results for these simulations showing the percentage of runs returning to the HHM are found in Fig 5. We found that no single intervention event allowed for a robust return from the AHM to the HHM indicating that two or more interventions are required. The least invasive treatment in the set of healthy outcomes included two of the treatable state variable nodes as targets. The best of these two-node treatments involved inhibiting GR and Th1Cyt. This two-target single intervention event produced a healthy outcome ~37% of the time in the 10,000 simulations. The single intervention treatment with the highest probability of delivering lasting remission involved four targets. To deliver a lasting remission with a maximum probability of ~57% required inhibiting ACTH, CORT, GR, and Th1Cyt. All single intervention courses resulting in a lasting return to HHM involved inhibition of GR, and all but one involved inhibition of Th1Cyt. This suggests that these two targets may constitute core treatment targets in more elaborate multiple intervention courses. The addition of additional interventions improved the overall probability of return to HHM however at the cost of invasiveness. Optimizing the Delivery of Multiple Intervention Events Analysis of the AHM basin of attraction indicates that multiple interventions are necessary to move the system to the HHM. As our discrete logic the system may be updated asynchronously, we use a genetic algorithm CAL 101 web coupled to our simulation to optimize the number, timing and type of interventions in a treatment course to move the system from the AHM to the HHM. Simulations were executed for treatment courses composed of 2 to 8 separate intervention events applied across multiple targets. These first two intervention events alone ended in stable and lasting return to the HHM in 40% of the simulated cases. Subsequent interventions consisted of additional treatment cycles alternating between repeated inhibition of Th1Cyt and/or GR. For example a second cycle of Th1Cyt and GR blockade produced a predicted remission rate of roughly 2 out of 3 simulated subjects. With 4 cycles of alternating Th1Cyt/ GR blockade separated by intermediate repeated suppression of GR delivered a maximum expected remission of rate of 86%. The intervals between intervention events varied between 10 and 50 time steps and depended on the specific sequence of treatment. Fig 6. GA Simulation results. doi:10.1371/journal.pone.0132774.g006 10 / 16 Achieving 
Remission in Gulf War Illness Simulating an Optimized Treatment Course PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19748594 Due to the limited availability of detailed kinetic data, the models used in this work strictly enforce the direction and sequence of regulatory control actions but do not represent the response time. Although the intervals described in Figs 6 and 7 are reported as simulation time steps, the actual timescale of these treatments courses can nonetheless be extracted in the form of landmark states. To provide a more robust approximation of the rate of success and to glimpse the intermediate transitory states occupied by the system, a minimally invasive single treatment cycle of Th1Cyt/ GR inhibition was simulated 1000 times. The value of each node variable was averaged PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19748686 for each time-point and plotted as a function of the time-step for all simulations leading to the HHM. These simulation results indicate that Th1Cyt inhibition alone causes t
