Fig 1.
Schematic illustrating pancreatic islets, circulation, and PLNs in the NOD mouse model.
Patches are shown by the background grids in NetLogo, turtles are mobile agents moving over the background grids (e.g. Naïve CD8+ T cell). DCs and CD8+ T cells migrate from circulation and/or PLNs to pancreatic islets, which contribute to T1D progression. One pancreatic islet is shown (the area restricted by the dashed curve in the pancreas). Pancreatic islets may overlap and form clusters of islets in the pancreas.
Fig 2.
Schematic diagram of T1D progression in circulation, islets, and PLN.
The legend shows the cell types and autoantigens used as agents in the ABM structure.
Fig 3.
Components of the agent-based model.
The diagram demonstrates all the interactions between agents in circulation, islets, and PLN and illustrates the level of intelligence in the ABM.
Table 1.
Data from experimental studies in the NOD mouse model.
Fig 4.
Individual-specific kinetics of infiltrating CD8+ T cells in T1D progression.
This figure shows a comparison of infiltrating CD8+ T cells between one hundred simulation runs and the experiment at week 4, week 6, week 8, week 10, week 12, and week 14, respectively. In this experiment, average counts of infiltrating CD8+ T cells (denoted by the red line) were calculated based upon 100 simulation runs (mean ± SEM and the shaded area represents variations in the simulation runs). Four representative trajectories of infiltrating CD8+ T cells (shown by dashed lines in blue, green, yellow, and cyan) reflect individual-specific patterns in the inherently noise-driven T1D progression. In addition, infiltrating CD8+ T cell counts (shown by the green square for mean values and black error bars for variations in the experiments) were illustrated using the raw data (7 replications at week 4 and week 12, 6 replications at week 6 and week 10, 5 replications at week 8, and 12 replications at week 14).
Fig 5.
Validation of the ABM model using experimental findings.
(A) Survival of healthy β cells during individual-specific T1D progression. Five solid lines represent the enduring fraction of healthy β cells in 100 distinct simulation runs with the same initial setting (ni = 20 and meani ± SEMi, for i∈{1: Blue, 2: Black, 3: Green, 4: Magenta, 5: Red}; parameter values can be found in S1 and S5 Tables). The vertical dotted-dashed line represents week 12 after the simulation initialization. The horizontal dotted-dashed line (y = 0.3) denotes the remaining 30% of healthy β cells, about which overt T1D may occur. (B) Distribution of overt T1D development occurring during week 12 and week 20 in 100 simulation runs corroborating experimental data [45,73,74]. The horizontal axis represents the time required for developing overt T1D in-silico, and the vertical axis denotes the relative frequency of T1D incidence. It is important to note that T1D onset occurs in our simulations based on the assumption that the survival percentage of healthy β cells falls randomly within the range of 10% - 30% of the initial healthy β cells (the total number of β cells at the simulation step 0).
Fig 6.
Discernible fluctuation in time required for developing overt T1D based upon selected values of parameters using local sensitivity analysis.
(A-E) The horizontal axes represent selected values of five input parameters in the spectrum of [Pd ± 30% × Pd], and the vertical axes denote the time required for developing T1D. Box plots and raw data illustrate heterogeneities occurring during T1D for selected values. The black dashed lines outline the lower and upper whiskers and black solid boxes show Q1, Q2, Q3 for the first quantile, interquartile, and third quantile values for selected points within the spectrum of [Pd ± 30% × Pd]. For these five sensitive parameters (shown by labels of the horizontal axes), the p-values of F tests were smaller than the predetermined significance level (e.g. α = 0.05).
Fig 7.
Change in model output based upon selected parameters.
(A) eFAST sensitivity based upon full data set; (B) eFAST sensitivity based upon reduced data set that includes only cases with T1D occurrence. The horizontal axes represent five sensitive parameters. The parameters P1, P2, P3, P4, P5 represent the average lifespan of cytotoxic CD8+ T cells in islets, the initial number of damaged β cells (βinit at simulation step 0), the time interval of DCs movement in islets, the maximum number of naïve CD8+ T cell binding to DCs, and recruitment rate of DCs circulating in the pancreas, respectively. The vertical axes denote the eFAST sensitivity analysis for the first-order and total-order indexes. The first-order indexes are denoted by pink portions starting from 0, and total-order indexes are illustrated by both grey and pink portions. The first-order index reflects the variance induced by the uncertainty of a single input factor and the total-order index represents the output variance abiding by the interaction between this input factor and other input factors.
Fig 8.
Probability of developing T1D based upon different targeted therapeutic strategies.
The horizontal axes represent the administered time of a single dosage during simulation runs. Regions with red color represent a higher incidence of diabetes, and regions with blue color represent a lower incidence of diabetes. (A) Therapy 1 denotes the strategy that reduces the longevity of CTLs in islets. (B) Therapy 2 is described as an intervention that inhibits DC infiltration into islets. (C) Therapy 3 denotes a strategy that mitigates the binding process of naïve CD8+ T cells on DCs. Therapeutic interventions were implemented based upon single dosage regimens starting at week 4 to week 16. Two black circles filled white in Panel (A) show the change in T1D occurrence in Therapy 1 when the lifespan of CTLs was reduced from 120 hours to 114 hours at week 8.
Fig 9.
Probability of developing T1D based upon combinations of proposed therapeutic strategies.
Therapeutic interventions were implemented based upon single dosage regimens starting at week 4 (Panels A1-A3) to week 16 (Panels D1-D3). Therapy 1 represents a certain strategy that reduces the residence of CTLs in islets. Therapy 2 is described as an intervention that can inhibit DC infiltration into islets. Therapy 3 denotes a strategy that prohibits binding sites on DCs for naïve CD8+ T cells.