GINI: From ISH Images to Gene Interaction Networks

Accurate inference of molecular and functional interactions among genes, especially in multicellular organisms such as Drosophila, often requires statistical analysis of correlations not only between the magnitudes of gene expressions, but also between their temporal-spatial patterns. The ISH (in-situ-hybridization)-based gene expression micro-imaging technology offers an effective approach to perform large-scale spatial-temporal profiling of whole-body mRNA abundance. However, analytical tools for discovering gene interactions from such data remain an open challenge due to various reasons, including difficulties in extracting canonical representations of gene activities from images, and in inference of statistically meaningful networks from such representations. In this paper, we present GINI, a machine learning system for inferring gene interaction networks from Drosophila embryonic ISH images. GINI builds on a computer-vision-inspired vector-space representation of the spatial pattern of gene expression in ISH images, enabled by our recently developed system; and a new multi-instance-kernel algorithm that learns a sparse Markov network model, in which, every gene (i.e., node) in the network is represented by a vector-valued spatial pattern rather than a scalar-valued gene intensity as in conventional approaches such as a Gaussian graphical model. By capturing the notion of spatial similarity of gene expression, and at the same time properly taking into account the presence of multiple images per gene via multi-instance kernels, GINI is well-positioned to infer statistically sound, and biologically meaningful gene interaction networks from image data. Using both synthetic data and a small manually curated data set, we demonstrate the effectiveness of our approach in network building. Furthermore, we report results on a large publicly available collection of Drosophila embryonic ISH images from the Berkeley Drosophila Genome Project, where GINI makes novel and interesting predictions of gene interactions. Software for GINI is available at http://sailing.cs.cmu.edu/Drosophila_ISH_images/     

A simulation study of the genetic regulatory hierarchy for butterfly eyespot focus determination

The color patterns on the wings of butterflies have been an important model system in evolutionary developmental biology. Two types of models have been used to study these patterns. The first type of model employs computational techniques and generalized mechanisms of pattern formation to make predictions about how color patterns will vary as parameters of the model are changed. These generalized mechanisms include diffusion gradient, reaction-diffusion, lateral inhibition, and threshold responses. The second type of model uses known genetic interactions from Drosophila melanogaster and patterns of candidate gene expression in one of several butterfly species (most often Junonia (Precis) coenia or Bicyclus anynana) to propose specific genetic regulatory hierarchies that appear to be involved in color pattern formation. This study combines these two approaches using computational techniques to test proposed genetic regulatory hierarchies for the determination of butterfly eyespot foci (also known as border ocelli foci). Two computer programs, STELLA 8.1 and Delphi 2.0, were used to simulate the determination of eyespot foci. Both programs revealed weaknesses in a genetic model previously proposed for eyespot focus determination. On the basis of these simulations, we propose two revised models for eyespot focus determination and identify components of the genetic regulatory hierarchy that are particularly sensitive to changes in model parameter values. These components may play a key role in the evolution of butterfly eyespots. Simulations like these may be useful tools for the study of other evolutionary developmental model systems and reveal similar sensitive components of the relevant genetic regulatory hierarchies.     

The relationship between early embryo development and tumourigenesis

With the recent substantial progress in developmental biology and cancer biology, the similarities between early embryo development and tumourigenesis, as well as the important interaction between tumours and embryos become better appreciated. In this paper, we review in detail the embryonic origin of tumour, and the similarities between early embryo development and tumourigenesis with respect to cell invasive behaviours, epigenetic regulation, gene expression, protein profiling and other important biological behaviours. Given an improved understanding of the relationship between early embryo development and tumourigenesis, now we have better and broader resources to attack cancer from the perspective of developmental biology and develop next generation of prognostic and therapeutic approaches for this devastating disease.     

Cloning, chromosomal organization and expression analysis of Neurl, the mouse homolog of Drosophila melanogaster neuralized gene

The Drosophila neuralized (neur) gene belongs to the neurogenic group of genes involved in regulating cell-cell interactions required for neural precursor development. neur mutant phenotypes include strong overcommitment to neural fates at the expense of epidermal fates. The human neuralized homolog (NEURL) has been recently determined and found to map to chromosome 10q25.1 within the region frequently deleted in malignant astrocytomas. Because of its potential importance in developmental processes, we analyzed the structure of the mouse homolog, Neurl, and its expression pattern in embryonic tissues. Neurl activity is detected from early developmental stages in several tissues and organs including neural tissues, limbs, the skeletal system, sense organs and internal organs undergoing epithelial-mesenchymal interactions. Neurl encodes a polypeptide associated with the plasma membrane but also detected in the cytoplasm. Similarly to the Drosophila gene, mammalian neuralized may code for an important regulatory factor.     

Automatic annotation of spatial expression patterns via sparse Bayesian factor models

Advances in reporters for gene expression have made it possible to document and quantify expression patterns in 2D-4D. In contrast to microarrays, which provide data for many genes but averaged and/or at low resolution, images reveal the high spatial dynamics of gene expression. Developing computational methods to compare, annotate, and model gene expression based on images is imperative, considering that available data are rapidly increasing. We have developed a sparse Bayesian factor analysis model in which the observed expression diversity of among a large set of high-dimensional images is modeled by a small number of hidden common factors. We apply this approach on embryonic expression patterns from a Drosophila RNA in situ image database, and show that the automatically inferred factors provide for a meaningful decomposition and represent common co-regulation or biological functions. The low-dimensional set of factor mixing weights is further used as features by a classifier to annotate expression patterns with functional categories. On human-curated annotations, our sparse approach reaches similar or better classification of expression patterns at different developmental stages, when compared to other automatic image annotation methods using thousands of hard-to-interpret features. Our study therefore outlines a general framework for large microscopy data sets, in which both the generative model itself, as well as its application for analysis tasks such as automated annotation, can provide insight into biological questions.     

Leg development in flies versus grasshoppers: differences in dpp expression do not lead to differences in the expression of downstream components of the leg patterning pathway

All insect legs are structurally similar, characterized by five primary segments. However, this final form is achieved in different ways. Primitively, the legs developed as direct outgrowths of the body wall, a condition retained in most insect species. In some groups, including the lineage containing the genus Drosophila, legs develop indirectly from imaginal discs. Our understanding of the molecular mechanisms regulating leg development is based largely on analysis of this derived mode of leg development in the species D. melanogaster. The current model for Drosophila leg development is divided into two phases, embryonic allocation and imaginal disc patterning, which are distinguished by interactions among the genes wingless (wg), decapentaplegic (dpp) and distalless (dll). In the allocation phase, dll is activated by wg but repressed by dpp. During imaginal disc patterning, dpp and wg cooperatively activate dll and also indirectly inhibit the nuclear localization of Extradenticle (Exd), which divide the leg into distal and proximal domains. In the grasshopper Schistocerca americana, the early expression pattern of dpp differs radically from the Drosophila pattern, suggesting that the genetic interactions that allocate the leg differ between the two species. Despite early differences in dpp expression, wg, Dll and Exd are expressed in similar patterns throughout the development of grasshopper and fly legs, suggesting that some aspects of proximodistal (P/D) patterning are evolutionarily conserved. We also detect differences in later dpp expression, which suggests that dpp likely plays a role in limb segmentation in Schistocerca, but not in Drosophila. The divergence in dpp expression is surprising given that all other comparative data on gene expression during insect leg development indicate that the molecular pathways regulating this process are conserved. However, it is consistent with the early divergence in developmental mode between fly and grasshopper limbs.     

Automated annotation of Drosophila gene expression patterns using a controlled vocabulary

MOTIVATION: Regulation of gene expression in space and time directs its localization to a specific subset of cells during development. Systematic determination of the spatiotemporal dynamics of gene expression plays an important role in understanding the regulatory networks driving development. An atlas for the gene expression patterns of fruit fly Drosophila melanogaster has been created by whole-mount in situ hybridization, and it documents the dynamic changes of gene expression pattern during Drosophila embryogenesis. The spatial and temporal patterns of gene expression are integrated by anatomical terms from a controlled vocabulary linking together intermediate tissues developed from one another. Currently, the terms are assigned to patterns manually. However, the number of patterns generated by high-throughput in situ hybridization is rapidly increasing. It is, therefore, tempting to approach this problem by employing computational methods. RESULTS: In this article, we present a novel computational framework for annotating gene expression patterns using a controlled vocabulary. In the currently available high-throughput data, annotation terms are assigned to groups of patterns rather than to individual images. We propose to extract invariant features from images, and construct pyramid match kernels to measure the similarity between sets of patterns. To exploit the complementary information conveyed by different features and incorporate the correlation among patterns sharing common structures, we propose efficient convex formulations to integrate the kernels derived from various features. The proposed framework is evaluated by comparing its annotation with that of human curators, and promising performance in terms of F1 score has been reported.     

Restricted spatial and temporal expression of G-protein alpha subunits during Drosophila embryogenesis.

Of the known signal transduction mechanisms, the most evolutionarily ancient is mediated by a family of heterotrimeric guanine nucleotide binding proteins or G proteins. In simple organisms, this form of sensory transduction is used exclusively to convey signals of developmental consequence. In metazoan organisms, however, the developmental role of G-protein-coupled sensory transduction has been more difficult to elucidate because of the wide variety of signals (peptides, small molecules, odorants, hormones, etc.) that use this form of sensory transduction. We have begun to examine the role of G-protein-coupled signaling during development by investigating the expression during Drosophila embryogenesis of a limited set of G proteins. Since these proteins are a common component of all G-protein-coupled signaling systems, their developmental pattern of expression should indicate when and where programmed changes in gene activity are initiated by, or involve the participation of, G-protein-coupled signaling events. We have focused on the spatial and temporal expression pattern of three different Drosophila G-protein alpha subunits by northern blot analysis, in situ hybridization and immunocytochemistry using antibodies directed to peptides specifically found in each alpha subunit. From the spatial and temporal restriction of the expression of each protein, our results suggest that different forms of G-protein-coupled sensory transduction may mediate developmental interactions during both early and late stages of embryogenesis and may participate in a variety of specific developmental processes such as the establishment of embryonic position, the ontogeny of the nervous system and organogenesis.     

Computational models for neurogenic gene expression in the Drosophila embryo

The early Drosophila embryo is emerging as a premiere model system for the computational analysis of gene regulation in development because most of the genes, and many of the associated regulatory DNAs, that control segmentation and gastrulation are known. The comprehensive elucidation of Drosophila gene networks provides an unprecedented opportunity to apply quantitative models to metazoan enhancers that govern complex patterns of gene expression during development. Models based on the fractional occupancy of defined DNA binding sites have been used to describe the regulation of the lac operon in E. coli and the lysis/lysogeny switch of phage lambda. Here, we apply similar models to enhancers regulated by the Dorsal gradient in the ventral neurogenic ectoderm (vNE) of the early Drosophila embryo. Quantitative models based on the fractional occupancy of Dorsal, Twist, and Snail binding sites raise the possibility that cooperative interactions among these regulatory proteins mediate subtle differences in the vNE expression patterns. Variations in cooperativity may be attributed to differences in the detailed linkage of Dorsal, Twist, and Snail binding sites in vNE enhancers. We propose that binding site occupancy is the key rate-limiting step for establishing localized patterns of gene expression in the early Drosophila embryo.     

The embryonic expression patterns of zfh-1 and zfh-2, two Drosophila genes encoding novel zinc-finger homeodomain proteins.

The zfh-1 and zfh-2 genes of D. melanogaster encode novel proteins containing both homeodomain and C2-H2 zinc-finger DNA-binding motifs. Antisera against these proteins were used to investigate their expression patterns during embryonic development. The zfh-1 gene is expressed in the mesoderm of early embryos and in a number of mesodermally-derived structures of late embryos, including the dorsal vessel, support cells of the gonads, and segment-specific arrays of adult muscle precursors. In addition, zfh-1 is expressed in the majority of identified motor neurons of the developing CNS. The mesodermal zfh-1 expression requires the products of the twist and snail genes. The zfh-2 gene displays a more limited expression pattern, largely restricted to the CNS of late embryos. Ubiquitous zfh-1 expression in transgenic flies bearing an hsp70-zfh-1 construct has specific developmental consequences, including embryonic CNS defects as well as adult eye and bristle abnormalities. The expression patterns of zfh-1 and zfh-2 suggest that both genes may be involved in Drosophila neurogenesis and that zfh-1 may have additional functions in mesoderm development.     

Constructing quantitative models from qualitative mutant phenotypes: preferences in selecting sensory organ precursors

MOTIVATION: To study biology from the systems level, mathematical models that describe the time-evolution of the system offer useful insights. Quantitative information is required for constructing such models, but such information is rarely provided. RESULTS: We propose a scheme-based on random searches over a parameter space, according to criteria set by qualitative experimental observations-for inferring quantitative parameters from qualitative experimental results. We used five mutant constraints to construct genetic network models for sensory organ precursor formation in Drosophila development. Most of the models were capable of generating expression patterns for the gene Enhancer of split that were compatible with experimental observations for wild type and two Notch mutants. We further examined factors differentiating the neural fate among cells in a proneural cluster, and found two opposite driving forces that bias the choice between middle cells and the peripheral cells. Therefore, it is possible to build numerical models from mutant screening and to study mechanisms behind the complicated network.     

Why we have (only) five fingers per hand: hox genes and the evolution of paired limbs.

Limb development has long been a model system for studying vertebrate pattern formation. The advent of molecular biology has allowed the identification of some of the key genes that regulate limb morphogenesis. One important class of such genes are the homeobox-containing, or Hox genes. Understanding of the roles these genes play in development additionally provides insights into the evolution of limb pattern. Hox gene expression patterns divide the embryonic limb bud into five sectors along the anterior/posterior axis. The expression of specific Hox genes in each domain specifies the developmental fate of that region. Because there are only five distinct Hox-encoded domains across the limb bud there is a developmental constraint prohibiting the evolution of more than five different types of digits. The expression patterns of Hox genes in modern embryonic limb buds also gives clues to the shape of the ancestral fin field from which the limb evolved, hence elucidating the evolution of the tetrapod limb.