Research Projects
Predictive modelling of C. elegans vulval development
C. elegans vulval development provides an important paradigm for studying the process of cell fate determination during animal development and it shares many characteristics with human biology. We are specifically interested in the crosstalk between the EGFR and LIN-12/Notch signalling pathways and how they orchestrate to control the process of pattern formation. We have previously constructed a dynamic, discrete, state-based model representing key aspects of cell fate specification during vulval development. The construction and execution of this model has highlighted important aspects of the biology of cell fate specification. All these aspects revolve around time/synchronicity issues: the timing of signal transduction and reception and creating a difference between fate decisions of initially equivalent cells.
The kind of discrete thinking used in the construction of such models is natural and intuitive; it is very suitable for the lack of quantitative data observed many times in biology. The type of high-level computer-aided reasoning is especially appropriate for the kind of models used by biologists to represent and reason about biological mechanisms, and could be applicable to many fields in biology. More recent work was based on the more sophisticated understanding of vulval fate specification that we have today. Formal analysis technique called model checking allowed us to test the consistency of the current conceptual model for vulval precursor cell fate specification with an extensive set of observed behaviours and experimental perturbations of the vulval system. The analysis of this model predicted new genetic interactions between the signalling pathways involved in the patterning process, together with temporal constraints that may further elucidate the mechanisms underlying precise pattern formation during animal development. These predictions were also validated experimentally in collaboration with Alex Hajnal (from the University of Zurich).
In addition, we are participating in a European consortium (PANACEA FP7) on quantitative pathway analysis of natural variation in complex disease signalling in C. elegans. The project focuses on collecting, analyzing and applying quantitative data to enable executable biology approaches addressing basic biological processes relevant to health; this is done in collaboration with the research groups of; Jan Kammenga (from Wageningen University), Gino Poulin (from University of Manchester), Alex Hajnal and Michael Hengartner (from University of Zürich), and Ritsert Jansen (from University of Groningen).
Computational modelling and analysis of the segmentation process in Drosophila embryogenesis
The Drosophila embryo is one of the developmental systems that have been studied extensively using classical genetic techniques. More recently the "Berkeley Drosophila Transcription Network Project" (BDTNP), has produced semi-quantitative 4D data on the behaviour of most of the important gene transcripts during a well-studied stage of development. Since Drosophila development has much in common with this of mammals, studying the process of segmentation is regarded as a possible Rosetta stone for deciphering human development.
The goal of this project is to build and analyze a computational model describing the segmentation process in Drosophila embryogenesis. The model (modules and variables) will be built from a description of the proposed gene regulatory network. The initial values and properties of interest will be determined from the BDTNP data. The analysis and discretization of the data available from the BDTNP database would serve as a test case for the analysis of similar databases. This work is done in collaboration with Angela DePace (from Harvard Medical School).
Mechanistic insights into metabolic disturbance in fat tissue during Diabetes and Obesity
Metabolic and inflammatory changes often observed in Diabetes and Obesity occur as a response to cellular stress, which includes oxidative stress, ER stress and hypoxia. The mechanistic origin and relative contribution of these stresses may differ between the acute and chronic situations. We have recently used the Qualitative Networks framework to model a metabolic network related to fat metabolism, which plays an important role in type-2 Diabetes and obesity. The model was built based on gene expression data obtained at different time points after a fat-feeding process. Analysis of the model has shown that MLXIPL plays a key role in this process as well as predicted new molecular interactions that were missing from the initial metabolic network. This work has suggested new experimental directions, which are now being checked at the lab of James Scott (from Imperial College London). We are currently interested to extend the model with the addition of genome-wide genetic and expression data and more quantitative metabolic data.
Modelling of the Notch/Wnt crosstalk in keratinocytes
The Notch, Wnt and EGFR signalling pathways are key players in the regulation of cell proliferation and differentiation and alterations in their function have been linked to several types of cancer. The aim of this project is to gain further insights into the role of Notch as a tumour suppressor in mammalian skin cells, possibly through its interaction with the Wnt and EGFR signalling pathways. Using the framework of Qualitative Networks we previously created a model describing the Notch/Wnt crosstalk during mammalian skin cells differentiation. Analysis of this model predicted that Jagged is a downstream target of Wnt signalling, a finding which was also validated experimentally. Currently we are aiming to extend this model with more recent molecular detail (e.g., P53, P63, EGFR signalling).
‘Computer programmed’ cell death
In collaboration with Adi Kimchi, David Harel, and Avital Sadot (from the Weizmann Institute) we study the process of programmed cell death (apoptosis) in a more systematic level. Kimchi’s lab had previously identified a group of pro-death proteins called DAPs (Death Associated Proteins) operating in the complex signalling network that regulates and executes various programmed cell death modules. By integrating this information gathered from various approaches (e.g., biochemistry, proteomics) into an executable model describing the process of apoptosis we hope to gain new insights into the molecular network underlying cell death.
Abstractions and large-scale modelling of biological signalling networks
Together with Luca Cardelli (from Microsoft Research Cambridge) we have previously created a pi-calculus model for the EGFR signalling pathway. Recently, we have been using this model to carry out simulations on systematically perturbed versions of the model in order to characterize the control functionality of each reaction involving key components in the signalling pathway. By partitioning the model into signalling modules, we were able to group control mechanisms and conduct model reduction. In future studies we will further investigated how best to simulates modules separately and combine the results of simulations.
Multi-scale modelling of animal development
Currently there are no tools that enable multi-scale modelling allowing to connect different levels of detail of the same biological system (e.g., molecular level, cellular level, organ level, whole-organism, etc.). The goal here is to create a unified environment allowing to construct, simulate, analyze, and visualize multiple models of different levels of detail (e.g., molecular interactions, cell-cell interactions, etc.) in the same platform, as well as to allow the communication between the different levels. The vision is that such a working environment could become a common practice in every lab, providing a user-friendly platform to formalize experimental data, execute working hypotheses and verify their consistency (automatically!) with the experimental data. We recently started to create such a multi-scale model of vulval precursor cells specification during C. elegans development, using pi-calculus and the SPiM tool to describe and execute the intracellular molecular interactions driving the intercellular signalling and cell fate acquisition specified in the Qualitative Networks framework. Live sequence charts (LSCs) will be used to verify the consistency of the model with experimental data. This is done in collaboration with Andrew Phillips and Hillel Kugler (from Microsoft Research Cambridge).
Formal modelling languages designed specifically for biological systems
The majority of existing modelling methods and tools were originally built for computerized systems. When modelling biological systems with such tools many features are redundant and others are missing. To illustrate this with a particular example, let’s consider parallel composition, which is the fundamental operation to build complex formal models from simple ones. In computer science, there are basically two different composition operators: synchronous and asynchronous. Biological systems are neither completely synchronous nor completely asynchronous. This is because different molecules, or cells, do not progress in perfect lock-step, but neither does any molecule or cell rest for arbitrary amounts of time. For this reason, we have previously introduced a new notion of “bounded asynchrony” into our models, which captures the phenomenon that biological systems proceed approximately along the same timeline.
Bounded asynchrony is but one example of how a formal modelling language designed specifically for biological systems might differ from modelling languages designed for hardware or software. Biological models do not need to be deployed but rather tested, simulated, and analyzed on the computer. This means that we can make choices that are impossible in the design of computerized systems. Redundant features that make analysis of computerized systems hard (or impossible) can be removed to enable stronger analysis. We aim to develop a new working environment (BioCharts) for simulation and analysis of biological models and we put a lot of emphasis on making this environment intuitive and user-friendly for biologists. We believe that such a modelling language will facilitate the use of executable models as a main stream technique in biological research.
In Qualitative networks we have extended Boolean networks by allowing elements to range over a small finite domain. Combining these with an iterative refinement and formal methods approach allow one to check the consistency of models with a set of laboratory experimental observations. This work was done in collaboration with Tom Henzinger (from the EPFL), Nir Piterman (from Imperial College London), and Marc Schaub (from Stanford).
