Blackboard teaching and Computer practicals
Qualitative dynamical modeling of (multi-) cellular networks
Claudine Chaouiya
Gulbenkian Institute, Lisbon
& Institut de Mathématiques de Marseille
We will start with the basics of the logical formalism to model and analyze regulatory/signaling networks. We will discuss a variety of computational methods that enable the analysis of rather large networks. The software GINsim (http://ginsim.org) will be used to build and analyze a simple logical model, illustrating the concepts introduced during the course. We will then introduce model composition and a cellular automata framework for the logical modeling of multi-cellular systems. In particular, we will focus on pattern formation in hexagonal grids, defining and simulating logical models using the software EpiLog (http://epilog-tool.org/).
Students need to have GINsim and EpiLog installed on their laptop, with a JAVA version 1.7 or higher. Files with dependencies should be downloaded from:
http://ginsim.org/sites/default/files/GINsim-3.0.0b-with-deps.jar
http://epilog-tool.org/sites/default/files/EpiLog-v1.1.1.jar
Dynamic models integrating metabolism and gene expression
Hidde de Jong
INRIA Grenoble – Rhône-Alpes
We will step-by-step build quantitative ODE models of a metabolic network integrating regulation on both the metabolic and gene expression level, and investigate the effect of these layers of regulation on the networks dynamics. The course will be structured around the case of carbon catabolite repression in bacteria, using the simple models described in Kremling et al. (2015). All simulations will be carried out by means of Matlab (no toolboxes needed), using code provided by the instructor.
Kremling A, Geiselmann J, Ropers D, de Jong H (2015). Understanding carbon catabolite repression in Escherichia coli using quantitative models. Trends Microbiol., 23:99-109
Resource Balance Analysis
Anne Goelzer
INRA Jouy en Josas
After a brief recall on constraint-based modeling, we will study the resource balance analysis (RBA) framework. We will illustrate the capability of prediction of RBA on a core-model of Bacillus subtilis: ribosome abundance, hierarchy of carbon and nitrogen source, cellular configurations maximizing the production of a compound of interest, etc. All simulations will be carried out by means of Matlab (release 2016 or later) and having either Cplex version 12.4.0.0 (preferred) or the optimization toolbox for solving linear programming optimization problems.
The RBA code using Cplex or the Matlab optimization toolbox can be found here. Please test it before the course by following these instructions.
https://www.ibm.com/analytics/cplex-optimizer
Stochastic modeling, simulation and analysis
Mustafa Khammash
ETHZ, Basel
The cellular environment is abuzz with noise. A key source of this noise is the randomness that characterizes the motion of cellular constituents at the molecular level. Cellular noise not only results in random fluctuations (over time) within individual cells, but it is also a main source of phenotypic variability among clonal cell populations. This course is concerned with the computational methods for modeling, simulation, and analysis of stochasticity in living cells.
The course topics include: introduction to stochastic gene expression; deterministic vs. stochastic models; the stochastic chemical kinetics framework; the chemical master equation; moment computations; the linear noise approximation; Monte Carlo simulations; Gillespie’s Stochastic Simulation Algorithm (SSA) and its variants. Exercises will require Matlab or R.
Enzyme economy in metabolic models
Wolfram Liebermeister
INRA Jouy en Josas
In this blackboard course, I will address an economic aspect of microbial metabolism: the protein cost associated with metabolic fluxes. I show how this cost can be approximated based on enzyme kinetics and on the assumption of minimal enzyme investments, and we will discuss how enzyme costs per unit flux, once they are known, can be used to predict metabolic fluxes and how simplified cost functions for Flux Balance Analysis can be derived. By considering a partitioning of protein resources between ribosomes and metabolic enzymes, predictions about enzyme cost can be translated into cellular growth rates. At the end of the course, we study constraint-based models that consider a fine-grained partitioning of the protein budget into individual enzymes. Such models predict metabolic strategies and protein investments solely from metabolic network structure, from physical and physiological constraints (such as limited cell space, protein composition, and presumable catalytic rates of enzymes), and from an assumed drive for fast growth.
Prerequisites: very basic understanding of cell biology, enzyme kinetics, and mathematical optimality problems.
Noor E., Flamholz A., Bar-Even A., Davidi D., Milo R., Liebermeister W. (2016). The protein cost of metabolic fluxes: prediction from enzymatic rate laws and cost minimization. PLoS Comput. Biol. 12: e1005167.
Wortel M.T., Noor E., Ferris M., Bruggeman F.J., Liebermeister W. (2018). Metabolic enzyme cost explains variable trade-offs between microbial growth rate and yield. PLoS Comput. Biol. 14: e1006010.
Simulation of intracellular mechanics
François Nédélec
EMBL Heidelberg
In this practical session, we will introduce basic methods to simulate systems containing mechanics and chemical reactions. We will in particular discuss how to extend Gillespie’s method when some of the reactions are affected by force, which is the case for example for the unbinding of a molecular bond under force, or for a molecular motor stepping along a filament. We will also discuss Brownian motion and consider simple problems of mechanical equilibrium, in which for example the position of a bead is constrained by a molecular link. The objective of the practical will be for the students to implement these methods and to write a simulation in Java of a bead being pulled by a molecular motor.
We will use the Processing software to be installed from https://processing.org/download/
Optimally learning dynamical models from data
Jakob Ruess
INRIA Saclay – Ile-de-France
& Pasteur Institute, Paris
In this blackboard course, we will focus on the identification of parameters of biochemical reaction network models (ODEs or stochastic models) from experimental data. Using a simple model of gene expression as case study, we will discuss how the amount of information about model parameters contained in different types of experimental data (e.g. population averages vs. single cell data) can be mathematically quantified. Subsequently, we will investigate how experiments can be optimally designed to maximize this information and how well chosen measurement times and/or perturbations of the system can help us to obtain precise estimates of parameter values.
Prerequisites: simple mathematical and statistical concepts such as (random) variables, probability distributions and ordinary differential equations.
