A model-based optimization framework for the inference of regulatory interactions using time-course DNA microarray expression data

Background: Proteins are the primary regulatory agents of transcription even though mRNA expression data alone, from systems like DNA microarrays, are widely used. In addition, the regulation process in genetic systems is inherently non-linear in nature, and most studies employ a time-course analysis of mRNA expression. These considerations should be taken into account in the development of methods for the inference of regulatory interactions in genetic networks. Results: We use an S-system based model for the transcription and translation process. We propose an optimizationbased regulatory network inference approach that uses time-varying data from DNA microarray analysis. Currently, this seems to be the only model-based method that can be used for the analysis of time-course "relative" expressions (expression ratios). We perform an analysis of the dynamic behavior of the system when the number of experimental samples available is varied, when there are different levels of noise in the data and when there are genes that are not considered by the experimenter. Our studies show that the principal factor affecting the ability of a method to infer interactions correctly is the similarity in the time profiles of some or all the genes. The less similar the profiles are to each other the easier it is to infer the interactions. We propose a heuristic method for resolving networks and show that it displays reasonable performance on a synthetic network. Finally, we validate our approach using real experimental data for a chosen subset of genes involved in the sporulation cascade of Bacillus anthracis. We show that the method captures most of the important known interactions between the chosen genes. Conclusion: The performance of any inference method for regulatory interactions between genes depends on the noise in the data, the existence of unknown genes affecting the network genes, and the similarity in the time profiles of some or all genes. Though subject to these issues, the inference method proposed in this paper would be useful because of its ability to infer important interactions, the fact that it can be used with time-course DNA microarray data and because it is based on a non-linear model of the process that explicitly accounts for the regulatory role of proteins.

A model-based optimization framework for the inference on gene regulatory networks from DNA array data

Vassily Hatzimanikatis, Eleftherios Papoutsakis

Identification of the regulatory structures in genetic networks and the formulation of mechanistic models in the form of wiring diagrams is one of the significant objectives of expression profiling using DNA microarray technologies and it requires the development and application of identification frameworks. Authors have developed a novel optimization framework for identifying regulation in a genetic network using the S-system modeling formalism. It has been shown that balance equations on both mRNA and protein species led to a formulation suitable for analyzing DNA-microarray data whereby protein concns. have been eliminated and only mRNA relative concns. are retained. Using this formulation, authors examd. if it is possible to infer a set of possible genetic regulatory networks consistent with obsd. mRNA expression patterns. Two origins of changes in mRNA expression patterns were considered. One derives from changes in the biophys. properties of the system that alter the mol.-interaction kinetics and/or message stability. The second is due to gene knock-outs. Authors reduced the identification problem to an optimization problem (of the so-called mixed-integer non-linear programming class) and developed an algorithmic procedure for solving this optimization problem. Using simulated data generated by our math. model, it has been shown that the method can actually find the regulatory network from which the data were generated. Authors also show that the no. of possible alternate genetic regulatory networks depends on the size of the dataset (i.e. no. of expts.), but this dependence is different for each of the two types of problems considered, and that a unique soln. requires fewer datasets than previously estd. in the literature. This is the first method that also allows the identification of every possible regulatory network that could explain the data, when the no. of expts. does not allow identification of unique regulatory structure. [on SciFinder (R)]

2004

Development and application of experimental tools to elucidate the gene regulatory network underlying adipogenesis

Carine Delattre-Gubelmann

Excess fat mass in obese individuals, which is characterized by an increase of white adipocyte cell size and number, significantly raises the risk of developing metabolic syndrome symptoms as well as cancer. A detailed understanding of the transcriptional mechanisms mediating white adipocyte differentiation (i.e., adipogenesis) is required to counteract the growing obesity epidemics. In this thesis, we have implemented two combinatory approaches to detect new protein-DNA interactions in the mouse, one of which targets the adipogenic gene regulatory network (GRN) that underlies the white adipocyte differentiation process. As a first approach to shed light on GRNs, we have developed and validated a powerful approach that combines a high-throughput yeast one-hybrid-based (Y1H) system to detect transcription factors (TFs)-DNA element interactions, together with a novel microfluidics-based technique, enabling the identification and characterization of individual binding sites for detected TFs within the respective regulatory sequences at unprecedented throughput and resolution. Using well-described regulatory elements as well as an uncharacterized enhancer which can play also a role in adipogenesis, we show that our approach in a first phase uncovers many known as well as novel TF-DNA interactions, of which a large proportion were independently validated. In addition, we further demonstrate how our approach can be used in a second phase to characterize detected interactions by enabling the generation of a relative DNA occupancy landscape over the length of the respective DNA element. Furthermore, we provide evidence that this system enables the detection of novel interactions that may have in vivo relevance. Given the strict requirements for direct and well-characterized protein-DNA interaction data to generate and model gene regulatory networks, the presented two-tiered approach should be of great value to enhance our understanding of the regulatory principles underlying mouse gene expression. To identify transcription factors involved in white adipocyte differentiation, we overexpressed TFs in preadipocytes and quantified their effect on differentiation. To this end, we have created a mouse open-reading frame resource consisting of more than 750 fully sequence-verified TF clones. We transferred these TF clones to a Tet-On lentiviral vector and used them to infect white preadipocytes. Based on microscopy-assisted quantification of lipids after induction of differentiation, we identified putatively novel TF candidates that strongly enhanced differentiation based on the increased lipid accumulation, indicating a potential role in adipogenesis. Gene expression and ChIP-seq experiments have been performed for the three top candidates (ZEB1, ZFP30 and ZFP277) to further study their function within the adipogenic regulatory network. To assess whether the effect of these TFs on adipogenesis could be reproduced in vivo, murine adipose stromal-vascular fraction containing adipocyte precursor cells were implanted into mice fed a high-fat diet, after each of the three TF candidates were knockdown or overexpressed in these cells. This experiment and the RNA-seq analyses are currently being processed while this thesis is being written. Furthermore, due to the absence of quantitative real-time PCR (qPCR) primer designing tools that would allow for gene-specific expression profiling in a high-throughput fashion, we design a user-friendly platform: GETPrime. This platform uniquely combines and automates several features critical to optimal qPCR primer design for all annotated Homo sapiens, Mus musculus, Caenorhabditis elegans, Drosophila melanogaster and Danio rerio genes in the Ensembl database. GETPrime primers have been extensively validated experimentally, demonstrating their high quality as well as high transcript specificity in complex samples.

EPFL2013

Quantitative single-cell analysis of S. cerevisiae using a microfluidic live-cell imaging platform

Johannes Becker

Genome-wide manipulations and measurements have made huge progress over the last decades. In Saccharomyces cerevisiae, a well-studied eukaryotic model organism, homologous recombination allows for systematic deletion or alteration of a majority of its genes. Important products of these manipulation techniques are two libraries of modified strains: A deletion library consisting of all viable knockout mutants, and a GFP library in which 4159 proteins are successfully tagged with GFP. In addition, the development of a method that allows for the systematic construction of double mutants led to a virtually infinite number of potential strains of interest. These advancements in combinatorial biology need to be matched by methods of data measurement and analysis. In order to simultaneously observe the spatio-temporal dynamics of thousands of strains from the GFP library, Dénervaud et al. developed a microfluidic platform that allows for parallel imaging of 1152 strains in a single experiment. On this platform, strains can be grown and monitored in a controllable environment for several days, which results in the imaging of several millions of cells during one experiment. To objectively and quantitatively analyze this immense amount of information, we implemented an image analysis pipeline, which can extract experiment-wide information on single-cell protein abundance and subcellular localization. The construction of a supervised classifier to quantify localization information on a single cell level is a new approach and was invaluable to detect dynamic localization changes within the proteome. Using five different stress conditions, we gained insight into temporal changes of abundance and localization of multiple proteins. For example, we found that while localization changes can often be fast and transient, long-term response of a cell is usually enabled by changes in abundance. This shows a well-orchestrated response of a cell to external stimuli. To extend knowledge about cellular mechanisms, we used our microfluidic platform for two separate screens, combining GFP-reporter with additional deletion mutants. The advantage of our platform in comparison to more common approaches lies in its simultaneous measurement of fluorescence and phenotypic information on cell size and growth. For each deletion, we can quantify not only its influence onto the respective GFP-reporter under changing conditions, but also its effect on cell growth and size. We showed that it is advantageous to combine this information, as it allows pointing out possible underlying mechanisms of gene network regulations. In a first screen we investigated the behavior of several gene networks upon UV irradiation damage. We were able to show that four gene deletions influenced the localization of ribonucleotide-diphosphate reductase (Rnr4p). A second screen was designed to find genes that influence the induction of the galactose network. This screen uses more than 500 deletions of genes mostly related to chromatin in combination with two different reporter strains. A main focus of this study was the inheritance of memory during galactose reinduction. We found several previously unknown genes that potentially influence either induction or reinduction and were picked as candidates for further inheritance studies. Our microfluidic platform allows for unprecedented studies of proteomes in flux. [...]

EPFL2015