Mapping Signaling Mechanisms in Neurotoxic Injury from Sparsely Sampled Data Using a Constraint Satisfaction Framework

Gulf War Illness (GWI) is a poorly understood exposure-induced neuroinflammatory disorder where complexity and the high cost of animal exposure studies has led to fragmented and sparse data sets incompatible with conventional data mining. We propose a numerical approach for generating hypotheses from sparse data to describe dysregulation of phosphoproteomic signaling in GWI brain. In an established animal model, hippocampus, and prefrontal cortex (PFC) samples were collected in mice exposed to corticosterone (CORT) to mimic high physiological stress, sarin surrogate diisopropyl fluorophosphate (DFP), CORT and DFP (CORT + DFP), as well as controls. IonStar liquid chromatography/mass spectrometry (LC/MS) profiling produced a network of 93 undirected interactions (Pearson correlation Bonferroni < 1%) linking 12 hippocampal and 5 PFC phosphoproteins. With only one pre-treatment resting state and one post-treatment transient observation, conventional rate models were infeasible. Instead, a simple discrete state transition logic was applied to each network node requiring baseline be a steady state from which the network could evolve through the transient 6-h post-treatment state. Solving this as a Constraint Satisfaction (SAT) problem produced 3 competing network models where DFP directly targeted phosphorylated subspecies of sodium channel protein type 1 subunit alpha (Scn1a), protein kinase C gamma (Prkcg), sacsin molecular chaperone (Sacs), in PFC and R3H domain containing 2 (R3hdm2) in hippocampus potentiated by corticosteroids. In simulation-based searches for intervention targets inhibition of Prkcg was disproportionately represented in rescuing the model-predicted persistent illness state, though companion targets were also necessary. Results such as these suggest that a dynamically constrained model-informed design can be highly useful in the initial phases of investigation into complex poorly understood illness where detailed data is largely unavailable.