An iterative pipeline of computational modelling and experimental design for uncovering gene regulatory networks in vertebrates

The behaviour of biological systems is the result of multiple regulatory interactions. Gene regulatory networks (GRNs) are the representation of these complex molecular interactions. They help us to understand how relationships between molecules dictate cell behaviour and are particularly useful to understand the complex dynamical processes driving animal development. The nodes in a GRN represent genes. Links between genes determine which gene products (proteins) regulate which other genes. In this project we focus on transcriptional regulation. This control mechanism regulates which genes are transcribed and expressed in the cell (and eventually which proteins are synthesized). The gene products that regulate transcription are a class of proteins called transcription factors. Transcription factors bind the DNA of target genes and influence the rate at which the target genes are transcribed. Mathematical models can describe how the rate of target gene transcription is affected when transcription factors bind to the DNA, and are useful to understand how cellular-scale behaviours arise from molecular actions. In this project we aim to develop computational methods able to construct GRN models using experimental data that describe gene expression and experimental data that describe the location of bound transcription factor proteins in DNA. An important aspect of the project is the development of a methodology capable of iteratively improving the GRN model by designing the most informative and effective sequence of experiments to be performed. This is particularly important since the experiments to locate binding of transcription factors to DNA are time-consuming and expensive and care should be taken to choose the most useful experiment at each stage. The methodology developed will be used to construct the GRN controlling the development of the second branchial arch (IIBA) in mouse. Branchial arches are transient structures of all vertebrate embryos that will eventually contribute to the face and the neck. Development of the IIBA is controlled by Hoxa2, a member of the large family of Hox transcription factors (TF). Hox TFs regulate morphogenesis along the head-tail axis of all animals with bilateral symmetry, but their mechanism of action is mostly unknown in vertebrates. Uncovering the principles underlying the GRN responsible for IIBA development will help us to understand the function of many other systems that are controlled by Hox proteins in vertebrate embryogenesis.

Gene regulatory networks (GRNs) determine the developmental program of organisms by controlling the dynamics of gene expression. Comprehensive GRN models are only available for the most experimentally tractable model organisms. Inferring GRN models in more complex organisms is hugely challenging because the availability of informative experimental perturbations is limited. Moreover, computational approaches to GRN inference assume the availability of well sampled expression data that are impractical to collect for most developmental systems. We propose an iterative pipeline of modelling and experiment to construct GRN models by integrating expression data with ChIP-seq data. Our methodology builds on a Gaussian process inference approach developed by applicants Rattray and Lawrence for inference in driven differential equation systems. We will integrate ChIP-seq experiments for the top-level transcription factors in the GRN with expression data (microarray and in situ hybridization from wild-type and mutant embryos) to determine a confident network scaffold. We will then construct a weighted ensemble of GRN models consistent with this scaffold using Bayesian methods. The ensemble will be iteratively refined using the principles of Bayesian experimental design to select the most informative additional ChIP-seq and in situ hybridization experiments. Our methodology will be used to model the GRN controlling an important developmental system in vertebrate embryogenesis, the second branchial arch (IIBA). The GRN that mediates IIBA development is controlled by one of the Hox transcription factors, Hoxa2. Applicants Bobola and Mankoo have carried out seminal work on Hoxa2 and Meox1, the key transcription factors in this system, and are therefore uniquely well placed to generate experimental data for modelling this system. Uncovering the principles underlying this GRN will help us understand other systems controlled by Hox proteins in vertebrate embryogenesis.

Improving the efficiency of the biosciences: The present project seeks to use Systems Biology models to direct further biological experiments in the most efficient and cost-effective manner. We predict that such a model-based experimental design approach will lead to great savings in various bio-medical projects and applications. The newest experimental procedures are always expensive initially (e.g. the ChIP-seq experiments considered here) and efficient experimental designs can potentially save large amounts of time and money, giving a competitive edge to industrial and academic researchers. Providing useful computational tools and improved data standards: Computational methodology developed in this project will be made available to a broad range of researchers, in both academic and industrial research environments. Rattray and Lawrence have an excellent track-record of making useful software available e.g. the open source puma package for microarray analysis, developed as part of a previous BBSRC-funded project, has been released through the Bioconductor project. In addition, by engaging with the SBML community on the inclusion of Bayesian methods we will increase the range of models that will be accessible through a standardised model format. Understanding complexity: The use of Systems Biology is also a consequence of our challenge to describe the high complexity of biological systems, organized in multiple hierarchical levels. The understanding of these systems cannot be explained solely on the basis of their individual parts; the resulting network organizations are typically robust, such that many single perturbations will not greatly affect them. The lessons learned from Systems Biology will feed into our understanding of complex systems in general, e.g. the environment, weather systems, energy networks, the economy, social networks and the world-wide web. Communications, Engagement and Training: The investigators in this proposal have an excellent track record of organising workshops and thematic programmes to engage with researchers across discipline boundaries, e.g. Lawrence and Rattray have recently edited a book entitled 'Learning and Inference in Computational Systems Biology' (MIT Press, in press) as a consequence of these activities. The project will provide excellent training in inter-disciplinary research for two Post-doctoral Research Assistants and associated PhD students from Manchester's Centre for Integrative Systems Biology. This will help to strengthen the UK's industrial and academic skills-base in a technical area that requires growth. As well as disseminating research and software through the standard outlets (journals, conferences, personal and project websites) we will interface with industry by presenting work directly to industrial contacts. The School of Computer Science at the University of Manchester runs an annual Industrial Forum. The idea behind the Forum is to introduce the School of Computer Science and all its myriad activities to industrial partners. The forum is made up of members of the School's industrial advisory board. Each year the School hosts a meeting of the forum that includes a poster exhibition and drinks reception. The posters reflect research being carried out in the School and include demonstrations. Any media relations activities and outreach will be through the University's media relations office, which is funded through the FEC component of the grant.