Computational Biology of Embryonic Stem Cells

Embryonic stem cells (ESC) have potential to be used in future therapeutic applications due to their unlimited self-renewal capabilities coupled with their ability to differentiate into any cell type. Mathematical models of ESC have gained much attention in recent years for their ability to extract information and insight from this self-renewal and differentiation system which might be otherwise elusive when using experimental data alone. In this chapter, we first present a brief review of previous efforts to model the ESC system, including foci on single cells, populations, self-renewal and differentiation mechanisms, and signaling and gene regulatory networks (GRN). GRN identification in ESC is invaluable, as proper information on network connections can give insight on how stem cells differentiate and can help in the development of efficient differentiation to specific cellular phenotypes. Although there has been considerable work on network identification of bacteria and the ESC self-renewal circuitry, work is still limited on differentiation. We therefore present our work on reverse engineering the gene regulatory network in differentiating ESC. In our network identification algorithm, we incorporate the inherent biological feature of sparsity, the notion that a network favors as few connections as possible. Our algorithm consists of a bi-level formulation, in which the upper level predicts the network topology and minimizes the number of connections, while the bottom level estimates the kinetic parameters and minimizes the error between predicted and experimental profiles. We apply our bi-level formulation to the system of mouse ESC differentiating towards pancreatic lineage. The input to the algorithm was the expression dynamics of relevant transcription factors. We show that the predicted gene behavior is in very good agreement with the in vitro experimental data, and that many of the interactions in the reconstructed network and predicted effects of external perturbations have been reported in literature, even though this information was not used to train the model a priori. The predictive capability of the algorithm was further substantiated by modeling the effect of Foxa2 silencing on differentiation outcome and validating it experimentally by gene silencing experiments.

In the past decade, technical innovations in systems biology has made it possible to study the activity of genes, pathways, transcription factors, metabolites and epigenetic states in vitro or in vivo on the genome-wide scale. In the design and analysis of such experiments, researchers often face an imperative question: To what extent and by which means can we extract valuable biological knowledge (which is often embodied as undirected and directed interactions between biological factors) from a particular experiment? In this chapter, we review state of the art algorithms for the structure learning of Bayesian networks and the elucidation of causal knowledge to partially address this question. Specifically, the distinct feature of each algorithm and its connections with other algorithms are highlighted in the context of causality reasoning and discovery for systems biology investigations.

Stem cell laboratories around the world routinely generate whole-genome expression data to study systems-level processes in stem cell biology, and computational clustering methods are critical for the genome-wide analysis of such large data sets. To address major limitations with commonly used clustering approaches, we developed a novel computational method called AutoSOME to automatically cluster large, high-dimensional data sets, such as whole-genome microarray expression data, without prior assumptions about cluster number or data structure. In previous work we demonstrated that AutoSOME clustering is an effective method for studying genome-wide expression patterns in stem cells. Here we present a primer that describes how to use this method to perform comprehensive cluster analyses of stem cell gene expression data. We include two detailed protocols illustrating the identification of gene co-expression modules and clusters of cellular phenotypes in a single step (Protocol 1), and the visualization of transcriptome variation among stem cells using an intuitive network display (Protocol 2). The workflow described in this chapter is sufficiently general for use with a wide variety of in-house and publicly available genomics data sets.

The promise of stem cell-based therapy is predicated on harnessing the plasticity of stem cell phenotypes to repair or replace damaged tissues. As technologies for detecting, isolating, modifying, tracking, and even inducing stem cells improve, the very definition of what constitutes a stem cell is now an open question. Addressing this fundamental problem has triggered an explosion of activity that spans the entire breadth of biological fields, from molecular biology to population biology. While this has clearly increased the gross amount of information concerning stem cells, its net impact is limited by a lack of integrative multiscale models that are readily accessible to researchers from many disciplines. The field of embryonic stem (ES) cell biology is a good example of the strengths and limitations of the segregative reductionist approach. The goal of this brief review is to highlight some of the most promising recent advances in embryonic stem cell research, with an emphasis on how data gathered from one level can benefit research across multiple scales.

Methylation of cytosines is a reversible and dynamic epigenetic DNA modification emerging during differentiation of human embryonic stem cells (hESCs) and throughout mammalian development. Crucial advancements in sequencing technologies have enabled the analysis of DNA methylation on a full genome level. Several studies recently examined the methylomes of hESCs, and investigated genetic and epigenetic dependencies during early differentiation. Methylated DNA immunoprecipitation (MeDIP) followed by high-throughput sequencing (MeDIP-seq) has become a cost-efficient experimental approach for genome wide epigenetic studies. However, it has been shown that MeDIP-seq data has to be corrected for a DNA sequence composition dependent bias in order to produce valid methylation profiles. Therefore, the development and implementation of time-efficient computational methods able to process large amounts of sequencing data with respect to its inherent complexity, is crucial for reducing the imbalance of sequencing data generation and analysis. This chapter introduces to different experimental techniques available for full genome methylation analysis. Subsequently, time efficient algorithms for processing MeDIP-seq data as well as different concepts for normalization are presented. Finally, recent findings of genetic and epigenetic dependencies in hESCs are summarized.

The decision between differentiation and self-renewal of embryonic stem cells (ESCs) are controlled by a complex network formed by interacting genes or proteins. Identifying critical components of this regulatory network and exploring their behavior patterns is crucial towards understanding the underlying mechanisms controlling ESC differentiation and realizing their potentials in regenerative medicine. In this chapter, we describe the usage of co-expression analysis in identifying conserved and divergent genomic, transcriptomic, and network modules critical for the earliest stage of ESCs differentiation.

The 3’ untranslated region (3’UTR) of mRNA plays important roles in posttranscriptional control of gene expression. Over half of the human genes have multiple polyadenylation sites in 3’UTRs, leading to 3’UTR isoforms containing different cis elements. Alternative polyadenylation (APA) has been found to be dynamically regulated in different tissue types and under various cellular conditions. Embryonic stem (ES) cells have the ability to self-renew and differentiate into any cell type in the adult body. Posttranscriptional gene regulation through cis elements in 3’UTRs is increasingly found to be important for these functions. In addition, various methods have recently been developed to induce differentiated cells to ES-like cells, called induced pluripotent stem (iPS) cells. Here we show a computational method to examine regulation of 3’UTR by APA using DNA microarray data. We applied this method to ES cells and iPS cells derived from different cell types.

Alternative splicing has the ability to expand proteome diversity among different cell and tissue types, and various stages of differentiation and development. In higher eukaryotes, alternative pre-mRNA splicing is highly regulated by many cellular factors and only in the past decade has the magnitude of this co/post-transcriptional regulation been revealed. Here we review recent technologies that have enabled genome-wide detection of alternative splicing, and highlight computational methods that have been developed to identify and understand the interplay between the cis and trans factors important for regulating alternative splicing in stem cells.

Spectacular advances in technology and computational biology spawned in large part by the Human Genome Project have been instrumental in transforming the field of embryonic stem cell biology. From the findings reported in the last decade it is clear that properties unique to embryonic stem cells (ESCs) are regulated not by individual genes but by complex gene networks that include both genes and ~22 nt noncoding microRNAs that act to integrate multiple genes across diverse signaling pathways to regulate self-renewal and differentiation. In this chapter we will discuss the evolution of our understanding of regulatory networks underlying stem cell self-renewal and pluripotency made possible through highthroughput genomic studies. In the last decade molecular technologies that revealed key transcription and epigenetic factors in ESCs have given way to highthroughput microarray and Next Generation Sequencing technologies. These largescale genomics datasets analyzed through the latest bioinformatic and computational methods have been instrumental in transforming the field of embryonic stem cells. We will trace the history of ES cells to briefly discuss key genes and microRNAs that have been established to regulate self-renewal and pluripotency in mouse and human prior to the genomics revolution. We will then discuss the latest technologies and computational algorithms that have been instrumental in revealing genome-wide changes associated with self-renewal and differentiation at the genetic and epigenetic levels to yield the current systems-level understanding of embryonic stem cells.

This chapter describes a method for genome-wide identification of translationally regulated genes during embryonic stem cell differentiation using integrated transcriptome and translation state profiling. Previous attempts at identification of translationally regulated genes have focused on measuring the fractionation of the mRNA molecules in the translated and untranslated fractions without considering the transcriptional status of these genes. In this method we describe how integrating information about transcriptional and translational status of genes enables in identification of translationally regulated genes with much greater accuracy. This approach developed for microarrays can also be used for gene expression measured by next generation sequencing.

Serial Analysis of Gene Expression (SAGE) is a sequence-based measure of gene expression that provides quantitative information on the population of transcripts through the generation and counting of specific sequence tags. Many SAGE datasets are publicly available for analysis, constituting a valuable resource for the study of gene expression. These datasets contain tags that are not obviously derived from known transcripts and thus hint at the existence of a large number of novel transcripts; however, the prioritization of candidates for further experimental verification is difficult. Here we demonstrate a method to identify non-coding antisense transcripts which may be implicated in stem cell differentiation by combining SAGE data with gene expression data derived by a complementary method. We produced SAGE libraries and paired microarray gene expression data pre- and post-differentiation of three mouse stem cell types (embryonic, mammary and neural). We found 1,674 SAGE tags antisense to 1,351 protein coding genes. A majority of these antisense tags overlap the 3’UTRs of sense genes; their abundance correlates with the expression of the corresponding sense genes and appears to be tissue specific. We did not find significant association between the expression of these tags and alternative splicing. We measured the expression of three genes expressed in the mouse embryo (Zfp42/Rex1, Ywhag/14- 3-3g and Pspr1) and corresponding putative antisense transcripts by qPCR before and after differentiation of mESC. We conclude that it is possible to identify putative novel antisense transcripts with a potential role in ES cell differentiation by integrating data from existing SAGE libraries with expression data derived by a complementary method. All data used in this work are available from the Gene Expression Omnibus (GEO) and StemBase databases.

With the advent of ultra-high-throughput sequencing technologies, ChIP-seq is becoming the main stream for the genome-wide studies of transcription factor binding sites (TFBSs) and histone modification sites. Computational analysis of ChIP-seq data is important for ChIP-seq applications. In this chapter, we first give an overview of the stateof- the-art ChIP-seq analysis tools developed for predicting ChIP-enriched genomic sites. Next, we describe the methods employed in a comprehensive analysis on Chen et al.’s ChIP-seq dataset in mouse embryonic stem cells (mESC) [1]. These methods include the prediction of transcription factor binding peaks, as well as subsequent analysis procedures such as de novo motif-finding and the discovery of transcription factor co-localization. By this, we demonstrate how the computational approaches assist to achieve novel biological discoveries from large-scale ChIP-seq dataset.

The “Functional Genomics in Embryonic Stem Cells” (FunGenES) database is based on gene expression profiling data obtained in selected pluripotent mouse Embryonic Stem (ES) cell lines using Affymetrix Mouse 430 v.2 arrays. The interactive FunGenES database allows users to derive gene expression profiles for every transcript that is included in the microarrays, or perform a series of gene association studies to search for groups of co-expressed and thus possibly co-regulated genes during ES cell growth and differentiation. It also includes advanced annotation tools and numerous connections to external reference tools and databases, linking gene expression patterns in stem cells to vital information about the role of corresponding genes in embryonic development or adult homeostasis and disease. In this chapter, I have used specific gene query examples to highlight the potential of the FunGenES database as a reference and discovery tool to study the biology of stem cells.