Computational Biology in Systems Immunology and Infectious Disease Modeling

点击翻译按钮获取中文摘要

Modern technology now allows the analysis of immune responses and host-pathogen interactions at a global level, across scales ranging from intracellular signaling networks, to individual cell behavior, to the functioning of a tissue, organ, and even the whole organism. The challenge is not only to collect the large amounts of data such methods permit, but also to organize the information in a manner that enhances our understanding of how the immune system operates or pathogens affect their hosts. Quantitative computer simulations are gaining importance as valuable tools for probing the limits of our understanding of cellular behavior. A major roadblock on the way to successful computational modeling in cell biology has been that the translation of qualitative biological models into computational models required the intervention of engineers/mathematicians as interfaces between biological hypotheses and their theoretical and computational representations. The software being developed by the computational biology group of the Laboratory of Immune System Biology eliminates the necessity of having this translation done by a person and thereby reduces the risk of oversimplification of biological mechanisms or the loss of important details in the course of translation by a non-biologist. The software ("Simmune") offers an intuitive graphical interface combined with state-of-the-art simulation technology. We recently added a module that automatically translates pathway models based on bi-molecular interactions into network visualizations that interactively display information about the details of the underlying reactions, such as required phosphorylations or induced molecular state transformations. Additionally, our software makes it possible to create computer simulations that combine detailed biochemical representation of cellular signaling processes with the spatial resolution necessary to reproduce the effects of localized recruitment and organization of signaling components. We have created a database and database interface system that can couple those computational models to experimental data and externally generated proteomic information. Our simulation software has the capability to combine biochemically detailed models with simulations that include morphological cellular plasticity. This makes it possible explore the interplay between cellular signaling processes and morphological dynamics that are controlled by those signaling processes while at the same time having a potentially strong influence on them. The ability of our modeling approach to simulate this combination of biochemical and morphological dynamics is based on algorithms we developed that are capable of automatically generating computational representations of intracellular reaction-diffusion networks. The input data provided by the user of our software consist of specifications of interactions between molecular binding sites and the modifications the interacting molecules undergo as a result of the interaction. These inputs - for which our software offers an intuitive graphical interface - are automatically translated into reaction-diffusion networks that reflect the specific geometry of the simulated cells. When the cells change their morphologies in the course of a simulation, the networks can, again automatically, be adjusted to reflect the new cellular shapes. We also develop components for this software that permit exploring the behavior of computational models over a wide range of parameter values to test whether a given model can reproduce experimental data, such as dose-response measurements, when its parameters are constrained only by what are considered physiologically reasonable ranges. In contrast to the commonly held assumption that a computational model can reproduce any data when it contains more than a handful of parameters, we found that even quite comprehensive models built with only mechanistic molecular interactions frequently fail to reproduce experimental data sets when these are sufficiently rich with regard to their dynamical or dose-dependent features. To improve the possibilities for model exchange between different modeling efforts we contributed to the development of a new standard for encoding multi-component / multi-state molecular complexes in SBML (Systems Biology Markup Language). In order to generate the necessary quantitative data to support the type of mechanistic models of signaling processes that can be built with our modeling approach (Simmune) we are performing systematic quantitative experimental measurements of intracellular signaling processes. For example, we measure the time courses of the activation of kinases and transcription factors downstream of cytokine and Toll-like receptors. We perform such measurements for varying doses of receptor stimuli to test whether our models behave correctly over a broad range of stimulation strengths and also analyze the effects to crosstalk that can occur when several pathways are activated simultaneously. To address the question of how much information is encoded in the interactions between receptors and ligands, we have extended the capabilities of a tool for analyzing amino acid complementarities among antibodies and their targets. We can now use a detailed information-theoretical framework to classify receptor-ligand interactions, for instance for T-cell receptors and their contact regions on MHC molecules, including the presented peptides. Finally, we developed a highly efficient stochastic particle-based simulation algorithm that combines components from Brownian Dynamics approaches and Greens Function Reaction Dynamics to permit large time steps (for maximal efficiency) while maintaining a high degree of precision. Recently, we were able to extend the capabilities of this simulation approach to include stochastic particle motion on curved surfaces as models of cellular membranes.