Fig 1.
Workflow of the developed simulation algorithm.
A. An asynchronously cycling cell population (Gen 1) is initiated by sampling the initial conditions at random time points from a pool of single cell simulations run with growth factor stimulation (Gen 0). B. Upon the execution of each generation, detection of new cell division events (or lack thereof) within simulation time determines the creation or not of a next generation. C. (left and center) Cross-generational trajectory of observed ERK and AKT activity from a randomly-chosen single-cell lineage. Varying colors represent subsequent generations, starting from Gen 1. (right) In silico lineage tracing capability is demonstrated with a lineage dendrogram. Lines representing individual cells are labeled with generation and index.
Fig 2.
Lineage Resolved Simulations for Comparing Simulated and Experimental Cell Viability Assays.
A-C. Dose response simulations for an example drug (trametinib). Median (across simulation replicates) cell population dynamics for several doses (A) and cell population lineage dendrograms for specific doses: 0 uM (B) and 0.1 uM (C) are shown. D-G. Simulated dose responses measured in GR-value for four drugs compared experimental data. Error bars are standard error taken from original experimental data, or as calculated across simulation replicates (n=10).
Fig 3.
Simulation Analysis to Investigate Palbociclib Dose Response Discrepancy.
A. Simulated cell growth curves under control conditions for multiple replicates. The dark black line is the median which was used to estimate doubling time, when the initial cell number (100) doubled (200). B. The fractional progression through the cell cycle (see Methods) was estimated for the beginning 100 generation 1 cells. This was the point at which 0.1 uM palbociclib was administered. The division outcome for each cell was then determined for this current generation, and if it exists for the next generation. The probability of division occurring was empirically estimated from this collection of binary outcomes and then plotted. C. Cell lineage dendrogram for response to 0.1 uM palbociclib. Most cells divide once early, and then the response is cytostatic.
Fig 4.
Simulation Analysis to Investigate Neratinib Dose Response Discrepancy.
A. Lineage dendrograms under control (no drug) conditions without EGF but with insulin. There are multiple rapidly dividing cells. B. Dependence of the probability of cell division as a function of the rate constant controlling basal Ras activation. Simulations were done as in A, without EGF and with insulin. The baseline value for the rate constant in the current published version of the model is designated by the red diamond. C-F. Dose response curves as in Figure 2, except with the altered basal RasGTP activation rate constant from panel B (2x10−4 s−1).
Fig 5.
Rapidly Dividing Phenotype in Control and Trametinib-Treated Conditions.
A. Histogram displaying low (0-6) moderate (6-16) and high (16+) number of division events for simulated cells under control conditions. Cells (4,000) were compiled across 4 drugs (0 dose), 100 cells per replicate, 10 replicates. B. Histogram displaying low (0-6) moderate (6-16) and high (16+) number of division events for simulated cells under trametinib treatment conditions (0.03 nM). Cells (1,000) were compiled across one dose, 100 cells per replicate, 10 replicates. C. Setup of the hypothesis, relating initial conditions of Generation 1 mother cells to the eventual number of divisions arising from them. D. Principal component plot of the initial states matrix under control conditions. Points are sized by number of division events, with colors equivalent to panel A. E. Top 20 loadings of the second principal component under control conditions, and a cartoon schematic of where they fall along the pathway driving the cell cycle. F. Projection of the trametinib data set onto the principal components learned from the control dataset. Points are sized by number of division events, with colors equivalent to panel A.