Session D04: Hierarchical Models for Omics Data

Fibration symmetries of integrated bio-network in a hypergraph representation of E.coli.

The multi-layer structure of integrated bio-network in E.coli. spans from transcriptional regulatory networks to metabolic networks where biochemical reactions happen. The interaction between the two networks characterizes the information transfer between metabolites and transcription factors (TF) and helps build coherent cellular states. Previously, fibration symmetries were found in transcriptional regulatory networks and uncovered the building blocks of biological networks. Here, we consider a hypergraph representation to model the integrated network of metabolites, and TFs, where dynamical features of metabolism are retained. We then apply balance coloring algorithm to study the fibration symmetries. We find that strong/weak fibration symmetries propagate from the regulatory networks to the metabolic networks, and, in turn, gene-metabolite complexes carry information from the metabolism to strengthen or weaken the symmetries in regulatory networks. Furthermore, the complexes act as switches controlling the general cellular status.   

Functional analysis of metabolism: a global approach unravels fluxes and protein allocation in E. coli

The complexity of the metabolic networks supporting the cellular production of energy and biomass represents a challenge to the genome-scale analysis of fluxes and proteome allocation. We developed a novel framework that allows to disentangle flux and proteome components associated with the production of energy and of individual biosynthetic precursors. After integrating genome-scale models of metabolism for Escherichia coli with experimental data on fluxes and cellular composition across conditions, we assigned experimental protein abundances to function-specific protein shares, thus quantifying the protein burden associated with each metabolic activity. This approach provides the basis for a function-based coarse graining of the cellular proteome and leads to the formulation of predictive models of protein allocation.   

Mechanism-based disease similarity: with random walks starting from a set of disease genes, traveling through the protein-protein interaction network, and then back.

Diseases have been traditionally classified based on the similarity between the symptoms they cause. One may wonder, however, could two or more diseases with drastically different symptoms have the same or similar underlying causes? To answer this question, it is essential to have a measure of similarity between diseases that reflects the underpinning molecular interactions. Such a measure can then be used for mechanism-based disease classification and for construction of a molecular-based human disease network. To achieve these goals, a bioinformatic tool called DeCoaD, which uses both genomic and proteomic data, has been developed. DeCoaD in turn utilizes ITMProbe, another in-house bioinformatic tool that employs damped random walks to analyze the flow of information in protein-protein interaction networks. In this talk the similarity measure and classification algorithm used by DeCoaD are presented and the resulting human disease network is explored. It is shown that the answer to the aforementioned question is positive. For example, DeCoaD demonstrates a significant similarity between dilated cardiomyopathy, a heart muscle disease, and infant diabetes mellitus. Several other examples of such seemingly unrelated pairs that have significant pair-wise similarities according to DeCoaD and the supporting experimental evince for these similarities are given. Finally, the utility of DeCoaD and the disease network for creation of a metadatabase for diseases is discussed. Such a metadatabse can be used for discovery of possible new disease relationships and validation of such relationships.   

Soft coral network diversity

Quantifying the morphology and morphogenesis of branching biological structures has recently progressed in systems across taxa. Our group focused on finding principles that govern the simple branching structure of gorgonians, flat soft corals commonly known as sea fans. Gorgonians have unusually plastic network topologies that are dependent on external environmental forces rather than tight DNA-encrypted regulation. Because of the quasi-2D morphology, these networks are amenable to high-throughput image analysis. We systematically compiled an image database of gorgonian morphologies including extant species sampled from every gorgonian order and several morphological variants of individual species. We then developed metrics to quantify their morphologies using simple measurements, and constructed a model dependent on local rules of self-organization that recapitulates many extant morphologies and is independent of a rigid genetic program. We believe that directing our focus to not-well-studied network systems like that of gorgonian morphology can bring forth novel information on network optimization.   

Principles of gene regulation quantitatively connect DNA to RNA and proteins in bacteria

Bacteria allocate their proteome to cellular functions differently in different growth conditions. It is largely unknown how such allocation arises from known mechanisms of gene regulation while constrained by limited translation capacity and fixed protein density. Here, we performed absolute transcriptomic and proteomic analysis for E. coli across many conditions, obtaining a plethora of results on promoters and mRNAs characteristics that clash with conventional expectations: the majority of mRNAs exhibit similar translational efficiencies, while the promoter strengths are vastly different across genes. These characteristics prescribe two principles of gene regulation guiding bacteria to attain the desired protein allocation under global constraints: Total transcriptional output is tightly coordinated with ribosomal activity, and the concentrations of individual proteins are largely set by transcription. These two principles lead to a quantitative formulation of Central Dogma which unravels the complex relationship between gene regulatory activities and mRNA/protein concentrations across conditions. The knowledge obtained will be invaluable for accurately inferring gene regulatory interactions from ’omics data, as well as for guiding the design of genetic circuits for synthetic biology applications in E. coli and other organisms.   

Analytical Descriptions of Fundamental Constraints in Protein Synthesis and Microbial Growth

The search for fundamental principles underlying cellular growth has long been a goal of microbiology. Recent years have yielded theoretical and experimental support for a candidate principle: growth in nutrient replete conditions is limited by protein synthesis and determined by how ribosomes are allocated towards making different protein classes. We introduce a low-dimensional model which investigates this principle, generating novel analytical expressions that define key properties of microbial growth (e.g.  translation rate and ribosome content) and explore how they depend on major physiological parameters (e.g. maximal translation rate and metabolic output). We then quantitatively explore strategies cells may employ for establishing specific growth rates where ribosomal content is either fixed across conditions, tuned to maintain fast translation, or tuned to optimize growth rate given the metabolic output. We compare our predictions to measurements of ribosome content and average translation rates for both E. coli and S. cerevisiae.  Despite their evolutionary distance, they share the same strategy for sculpting their proteome; ribosome synthesis is tuned to promote fast growth at the expense of the average translation rate. The agreement between theory and experiment illustrates that fundamental biological principles are encoded in this model. We discuss how this is achieved in E. coli and comment on how these findings relate to the eco-evolutionary histories of these organisms.   

Bacterial adaptation to changing conditions – an eco-physiological case study with the gut bacterium Escherichia coli

To thrive, microbial organisms must successfully cope with the changing conditions they regularly encounter. The response to changing conditions typically requires the synthesis of new enzymes - for example a specific transporter to utilize a newly available nutrient source. But protein synthesis is resource demanding and novel protein synthesis by ribosomes is a fundamental constraint which limits growth even in  nutrient replete conditions. Cells must thus devise effective regulation strategies which control the synthesis of different proteins --  in relation to the environmental conditions they encounter and what physiology demands to promote survivival and growth. To better understand these regulation strategies and their eco-physiological origin, we have investigated the response of Escherichia coli to changes in available nutrient sources, connecting transient gene-expression and proteome composition to growth phenotypes. By combing experiments and theory we show how competition between genes for the limited protein synthesis capacity constrains growth during the transition. Despite this constraint, cells substantially express genes that are not required in the conditions they encountered, trapping them in states where precursor levels are low and the genes needed to replenish the precursors are outcompeted. These observations might explain why bacteria frequently show long phases of growth arrest when conditions change. Contrary to common modeling assumptions, our findings also highlight that microbial cells do not attempt to optimize growth under changing environments but rather exhibit very specific response strategies which likely evolved to promote fitness in the specific native environment cells typically encounter.    

Top-down identification of keystone species in the microbiome

Keystone species in ecological communities are native species that play an especially important role in the stability of their ecosystem and can also be potentially used as its main drivers. However, we still lack an effective framework for identifying these species from the available metagenomic data without the notoriously difficult step of reconstructing the detailed network of inter-specific interactions.

Here we propose a top-down identification framework, which detects keystones by their total influence on the rest of the species. Our method does not assume pairwise interactions or any specific underlying dynamics and is appropriate to both perturbation experiments and metagenomic cross-sectional surveys. When applied to real metagenomic data of the human gastrointestinal microbiome, we detect a set of candidate keystones and find that they are often part of a keystone module -- multiple candidate keystones species with correlated occurrence.  The keystones analysis of single-time-point cross-sectional data is also later verified by evaluation of two-time-points longitudinal sampling.

Our framework represents a necessary advancement towards the reliable identification of these key players of complex, real-world microbial communities.   

Withdrawn

Advances in time-lapse microscopy mean individual cells can be tracked as they move through the cell cycle and their lineage information obtained. In the literature, many interesting results have been revealed by analysing the correlation in interdivision time of various family pairs of cells. 

We introduce a general stochastic matrix model of abstract cell-cycle factors that are inherited from the mother to daughter cell. The model can be parameterised from measured correlations alone, and contains common models of cell cycle, replication and size control as special cases.

Using Bayesian inference, we fit the model to six lineage tree datasets of bacteria and mammalian cells from available literature. Our model gives a consistently good fit to the data, despite the model parameters being unidentifiable. This suggests that using interdivision time data alone is not sufficient to identify precise cell cycle control mechanisms. However, analysing the posterior distributions reveals oscillations driving the cell cycle. The frequencies of these oscillations can be attributed to underlying biological oscillators, such as circadian rhythm. This insight helps us to understand the factors that affect cell-to-cell variability in interdivision time.   

Shell Proteins' Roles in Bacterial Microcompartment Assembly

Bacterial microcompartments are nanometer-scale proteinaceous containers that house biochemical pathways in bacteria, protecting the organism from toxic intermediates and co-localizing enzymes. Recently, there has been much interest in re-engineering these systems for use as "metabolic modules" through the engineering of shell proteins. However, this can be a daunting task, since these shells can be composed of up to eight unique shell proteins whose roles in the assembly are not yet well understood. Here, we use atomistic simulations to measure a vast range of interaction strengths including a multitude of different bending interactions between pairs of Pdu shell proteins. These interactions inform a coarse grain model that shows how varying these interactions in a system with multiple types of shell proteins leads to different assembly paths and shells with different properties. Specifically, we show an assembly where one shell protein forms a nucleus in the bulk, which is then recruited to the enzymatic core by a second type. Once this nucleus binds the cargo a third type of shell protein helps to close the shell. These shells have, on average, weak interactions with the enzyme core leading the shell to be larger and partially empty. Computational results are experimentally validated throughout.