Calibrating agent-based models to tumor images using representation learning.

Agent-based models (ABMs) have enabled great advances in the study of tumor development and therapeutic response, allowing researchers to explore the spatiotemporal evolution of the tumor and its microenvironment. However, these models face serious drawbacks in the realm of parameterization - ABM parameters are typically set individually based on various data and literature sources, rather than through a rigorous parameter estimation approach. While ABMs can be fit to simple time-course data (such as tumor volume), that type of data loses the spatial information that is a defining feature of ABMs. While tumor images provide spatial information, it is exceedingly difficult to compare tumor images to ABM simulations beyond a qualitative visual comparison. Without a quantitative method of comparing the similarity of tumor images to ABM simulations, a rigorous parameter fitting is not possible. Here, we present a novel approach that applies neural networks to represent both tumor images and ABM simulations as low dimensional points, with the distance between points acting as a quantitative measure of difference between the two. This enables a quantitative comparison of tumor images and ABM simulations, where the distance between simulated and experimental images can be minimized using standard parameter-fitting algorithms. Here, we describe this method and present two examples to demonstrate the application of the approach to estimate parameters for two distinct ABMs. Overall, we provide a novel method to robustly estimate ABM parameters.

Proof-of-concept for incorporating mechanistic insights from multi-omics analyses of polymyxin B in combination with chloramphenicol against Klebsiella pneumoniae.

Carbapenemase-resistant Klebsiella pneumoniae (KP) resistant to multiple antibiotic classes necessitates optimized combination therapy. Our objective is to build a workflow leveraging omics and bacterial count data to identify antibiotic mechanisms that can be used to design and optimize combination regimens. For pharmacodynamic (PD) analysis, previously published static time-kill studies (J Antimicrob Chemother 70, 2015, 2589) were used with polymyxin B (PMB) and chloramphenicol (CHL) mono and combination therapy against three KP clinical isolates over 24 h. A mechanism-based model (MBM) was developed using time-kill data in S-ADAPT describing PMB-CHL PD activity against each isolate. Previously published results of PMB (1 mg/L continuous infusion) and CHL (Cmax : 8 mg/L; bolus q6h) mono and combination regimens were evaluated using an in vitro one-compartment dynamic infection model against a KP clinical isolate (108 CFU/ml inoculum) over 24 h to obtain bacterial samples for multi-omics analyses. The differentially expressed genes and metabolites in these bacterial samples served as input to develop a partial least squares regression (PLSR) in R that links PD responses with the multi-omics responses via a multi-omics pathway analysis. PMB efficacy was increased when combined with CHL, and the MBM described the observed PD well for all strains. The PLSR consisted of 29 omics inputs and predicted MBM PD response (R2  = 0.946). Our analysis found that CHL downregulated metabolites and genes pertinent to lipid A, hence limiting the emergence of PMB resistance. Our workflow linked insights from analysis of multi-omics data with MBM to identify biological mechanisms explaining observed PD activity in combination therapy.

Multi-scale modeling of macrophage-T cell interactions within the tumor microenvironment.

Within the tumor microenvironment, macrophages exist in an immunosuppressive state, preventing T cells from eliminating the tumor. Due to this, research is focusing on immunotherapies that specifically target macrophages in order to reduce their immunosuppressive capabilities and promote T cell function. In this study, we develop an agent-based model consisting of the interactions between macrophages, T cells, and tumor cells to determine how the immune response changes due to three macrophage-based immunotherapeutic strategies: macrophage depletion, recruitment inhibition, and macrophage reeducation. We find that reeducation, which converts the macrophages into an immune-promoting phenotype, is the most effective strategy and that the macrophage recruitment rate and tumor proliferation rate (tumor-specific properties) have large impacts on therapy efficacy. We also employ a novel method of using a neural network to reduce the computational complexity of an intracellular signaling mechanistic model.

Modeling immune cell behavior across scales in cancer.

Detailed, mechanistic models of immune cell behavior across multiple scales in the context of cancer provide clinically relevant insights needed to understand existing immunotherapies and develop more optimal treatment strategies. We highlight mechanistic models of immune cells and their ability to become activated and promote tumor cell killing. These models capture various aspects of immune cells: (a) single-cell behavior by predicting the dynamics of intracellular signaling networks in individual immune cells, (b) multicellular interactions between tumor and immune cells, and (c) multiscale dynamics across space and different levels of biological organization. Computational modeling is shown to provide detailed quantitative insight into immune cell behavior and immunotherapeutic strategies. However, there are gaps in the literature, and we suggest areas where additional modeling efforts should be focused to more prominently impact our understanding of the complexities of the immune system in the context of cancer. This article is categorized under: Biological Mechanisms > Cell Signaling Models of Systems Properties and Processes > Mechanistic Models Models of Systems Properties and Processes > Cellular Models.

Data-driven analysis of a mechanistic model of CAR T cell signaling predicts effects of cell-to-cell heterogeneity.

Due to the variability of protein expression, cells of the same population can exhibit different responses to stimuli. It is important to understand this heterogeneity at the individual level, as population averages mask these underlying differences. Using computational modeling, we can interrogate a system much more precisely than by using experiments alone, in order to learn how the expression of each protein affects a biological system. Here, we examine a mechanistic model of CAR T cell signaling, which connects receptor-antigen binding to MAPK activation, to determine intracellular modulations that can increase cellular response. CAR T cell cancer therapy involves removing a patient's T cells, modifying them to express engineered receptors that can bind to tumor-associated antigens to promote tumor cell killing, and then injecting the cells back into the patient. This population of cells, like all cell populations, would have heterogeneous protein expression, which could affect the efficacy of treatment. Thus, it is important to examine the effects of cell-to-cell heterogeneity. We first generated a dataset of simulated cell responses via Monte Carlo simulations of the mechanistic model, where the initial protein concentrations were randomly sampled. We analyzed the dataset using partial least-squares modeling to determine the relationships between protein expression and ERK phosphorylation, the output of the mechanistic model. Using this data-driven analysis, we found that only the expressions of proteins relating directly to the receptor and the MAPK cascade, the beginning and end of the network, respectively, are relevant to the cells' response. We also found, surprisingly, that increasing the amount of receptor present can actually inhibit the cell's ability to respond due to increasing the strength of negative feedback from phosphatases. Overall, we have combined data-driven and mechanistic modeling to generate detailed insight into CAR T cell signaling.