Initial overview of Human Quantitative Dynamics: A project to quantify the components of and interactions in a human cell
Ravi Iyengar
Department of Pharmacology and Systems Therapeutics and Systems Biology Center, New York, Icahn School of Medicine at Mount Sinai, New York, NY
Understanding the relationship between genotype and phenotype is a multistep process that requires deciphering the relationship between the genome and the transcriptome, control of mRNA levels by genomic determinants as well as epigenomic regulators and expression patterns of mRNA in the different cell types. Each of these relationships has been or is being mapped in projects supported by the NIH, such as ENCODE and GTeX. A necessary next step is to identify and quantify the levels and interactions of cellular components including proteins and key small molecules and metabolites that give rise to whole cell behaviors that are characteristic of different cell types. The Human Quantitative Dynamics Project will focus on identifying the components of human cells and mapping their interactions in a precise and quantitative manner. Such a goal will require a complex integrated enterprise that uses multiple types of technologies, some of which currently exist and some that need to be developed, and the engagement of different biomedical and engineering communities. In this talk, I will present a listing of both the opportunities and challenges designing such a project poses, and provide suggestions for potential approaches including mechanisms that can enable collaborations between researchers in high-throughput and focused research communities. I will also present an aspirational view of how the output of such a project would look and how it could drive biomedical research in the coming years.
9:20 am - What can HQD for Basic Research focused on understanding complex multi-scale biology like tissue and organ development and physiology/homeostasis Marian Walhout - University of Massachusetts Medical School
What can HQD do for basic research focused on understanding complex multiscale processes, such as development and homeostasis?
A.J. Marian Walhout
Program in Systems Biology, University of Massachusetts Medical School, Worcester, MA
Systems biology studies of development and homeostasis have heavily focused on measurements of the transcriptome, proteome and interactome in model organisms such as yeast, worms, and flies, and to a lesser extent in mammalian tissues and cells. However, quantitative models that describe how the transcriptome and interactome function and interact are lacking. We will discuss the types of data that are required for this next important challenge: to obtain predictive models that capture large-scale systems-level data to obtain deep mechanistic insights.
9:40 am - What can HQD do for Precision Medicine Garret FitzGerald - University of Pennsylvania
Department of Systems Pharmacology and Translational Therapeutics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA
Clinical research has traditionally rested on the ability of randomized, double blind, placebo-controlled trials to detect large, average effects of drug efficacy and safety, and, to a lesser extent, data from observational studies claimed to reflect, albeit less precisely, “real life” situations. Presently, it is undergoing a fundamental shift. The rapid spread of electronic health records permit one to seek at scale disease associations which might never have been thought of or detected in the context of an individual’s patient experience. Linked biobanks might be exploited to seek biomarker evidence consistent with mechanistic linkage. If this thesis is supported, attention might then turn to deep phenotyping in small numbers of individuals – human phenomic science – to seek proof of concept before performing more efficiently than hitherto, targeted, refined (e.g., adaptive) prospective trials in pursuit of information more precisely linked to the individual. Such precision medicine strategies require the integration over time of diverse data sets: omics, imaging, physiological outputs like blood pressure and other remote sensing outputs, such as what and when foodstuffs are consumed, activity, and communication patterns, for example. The objective may be to develop predictive paradigms of disease evolution across the lifespan, or of efficacy or hazard of therapeutic interventions. Either way, repeated sampling over time in ways that permit iterative refinement of the models are necessary. In the case of therapeutics, the documentation of actual drug consumption and then the accurate measurement of systemic, and ideally local, drug exposure must be integrated with outputs reflective of drug response. Dynamical modelling can be directly integrated with hypothesis testing in human induced pluripotent stem cells ex vivo, and indeed, with data obtained in model systems, further to refine predictive paradigms in clinical research.
10:00 am Coffee Break
Session 2 What sort of measurements do we need to fully understand quantitative dynamics ?
10:15 am - The Human Proteome Atlas Mathias Uhlen Royal Institute of Technology (KTH)
KTH Biotechnology, AlbaNova University Center, Royal Institute of Technology (KTH), Stockholm, Sweden
Modern biological and medicinal researchers increasingly rely on technologies for rapid and large-scale the comparative analysis of proteomes. Here, we describe the first identification of ~4,000 yeast proteins, a comprehensive yeast proteome, from just over an hour of analysis time. Next, we discuss the application of this technology to profile over one thousand proteomes and metabolomes for functional annotation of orphan genes. Prospects for extending this technology for ultrafast whole proteome quantification in human systems will also be presented. Finally, we describe technology for near complete mapping of the human proteome. With these methods we have achieved detection over 17,000 proteins in a single cell line and have detected peptides covering over 70% of each protein. Specifically, we recorded mass spectral evidence for 7 of every 10 amino acids in the human proteome. To our knowledge these data represent the deepest and most comprehensive proteomic analysis of the human proteome acquired to date.
10:35 am - Analysis of network perturbations in human disease Marc Vidal - Harvard Medical School
Analysis of network perturbations in human disease
Marc Vidal
Center for Cancer Systems Biology (CCSB), and Department of Cancer Biology, Dana-Farber Cancer Institute, and Department of Genetics, Harvard Medical School, Boston, MA
Our goal is to contextualize and functionalize human genetic variants for a wide range of disorders, from simple Mendelian diseases to complex traits, through direct interrogation of macromolecular networks. Along with network analysis of wild type and mutant proteins, we are developing concepts, technologies, and systematic datasets to functionally characterize large numbers of genotypes in terms of the effects they have on the molecular functions and physical and biochemical interactions mediated by the corresponding gene products, with implications for elucidating human gene-gene interactions and understanding heritability and improving genomic medicine. I will discuss how these approaches can be readily adapted to comprehensively map interactions within different human cell types.
10:55 am - The protein database landscape: Achievements and limitations Rolf Apweiler - EMBL-EBI Wellcome Trust
The protein database landscape: Achievements and limitations
Rolf Apweiler
EMBL-EBI Wellcome Trust Genome Campus, Hinxton, Cambridgeshire, UK
It is important to address informatics and database challenges if we want to move towards a detailed and comprehensive quantitative understanding of the components of various cell types and tissues in terms of their concentrations and kinetic interactions. In my talk, I will give my view on the current database landscape, from capturing of raw experimental data of a comprehensive quantitative project rooted in biochemistry and cell physiology to analysis and organization of the data into functionally annotated proteins, complexes and pathways. Finally, I will try to point out limitations in our current database landscape, and in our ability to integrate this knowledge with genomic and epigenomic information, and how to overcome these limitations.
11:15 am - Crowdsourcing approaches for parameter estimation and building whole cell models Gustavo Stolovitzky - IBM Research
Crowdsourcing approaches for parameter estimation and building whole cell
Gustavo Stolovitzky
Functional Genomics and Systems Biology, IBM Research, Thomas J Watson Research Center, Yorktown Heights, NY
Once the gene regulatory network (GRN) of a biological process has been well characterized, predicting its dynamics depends on accurately estimating the parameters governing its biochemical kinetics. It is then essential to choose the most informative experiments and numerical approaches that will help estimate the parameter values. In the context of the Dialogue for Reverse Engineering Assessments and Methods (DREAM), my colleagues and I created an in silico test framework under which participants could apply their algorithms to determine the unknown parameters of different types of proposed GRNs. To perform their parameter estimation, participants could use a limited set of data chosen from in silico experimental assays and perturbations derived from the GRN ODE model. This successful approach was later applied to a much larger set of models, the so-called whole-cell models, that explicitly represent all cellular components at the molecular level and may enable genotype-to-phenotype predictions. Whole-cell models typically have thousands of parameters, many of which are poorly characterized or unknown. New algorithms are needed to estimate these parameters and enable researchers to build increasingly comprehensive models; hence, we asked Challenge participants to identify a subset of parameters of a modified Mycoplasma genitalium whole-cell model given the model's structure and in silico "experimental" data. The results from these DREAM challenges have been very informative regarding the ability to estimate the parameters of a whole range of GRN, and I will describe several lessons in the design and execution of crowdsourced challenges, and how such community efforts could be integrated with large-scale data gathering efforts.
11.35 am - Formatting the Open Discussion - This session will be focused on developing questions and structure for after lunch Open- Structured-Discussion Brenda Andrews - University of Toronto
11:50 am Lunch
Discussion I – The Big Picture and Scope
12:45 pm – 1:30 pm - Brenda Andrews
1. Will HQD, if successful, be useful for biomedical sciences? What would be the benchmarks for success?
2. Proteins, metabolites, and interactions in a specific cell type: how complete is complete?
3. There is very little incentive to perform high throughput biochemical studies now. How do we encourage high throughput technology development in
biochemistry and cell biology to study dynamics?
Session 3 Successes and failures of large-scale projects: what can we learn to design new projects
An overview of trans-NIH funding opportunities for large scale scientific projects will be discussed. Topics will include the NIH Common Fund, NIGMS Glue Grants, and NIGMS Systems Biology Centers. A brief introduction to the other speakers in Session 3 will be provided.
1.40 pm - Multiomics projects for precision medicine: Dynamic analysis of healthy and disease states using personal omics profiling Mike Snyder - Stanford University
Department of Genetics, Stanford University, Stanford CA
Understanding health and disease requires a detailed analysis of both our DNA and the molecular events that determine human physiology. We performed an integrated Personal Omics Profiling (iPOP) on 100 healthy and prediabetic human subjects over periods of viral infection as well as during controlled weight gain and loss. Our iPOP integrates multiomics information from the host (genomics, epigenomics, transcriptomics, proteomics and metabolomics) and from the gut microbiome. Longitudinal multiomics profiling reveals extensive dynamic biomolecular changes occur during times of perturbation, and the different perturbations have distinct effects on different biomolecules in terms of the levels and duration of changes that occur. Overall, our results demonstrate a global and system-wide level of biochemical and cellular changes occur during environment exposures. I will discuss both the advantages and challenges of multiomics projects and how our experiences could be relevant for the proposed Human Quantitative Dynamics project.
2:00 pm - Studying genetic interaction networks Brenda Andrews - University of Toronto
The Donnelly Centre for Cellular and Biomolecular Research, and Department of Molecular Genetics, University of Toronto, Toronto, ON, Canada
We generated a comprehensive genetic interaction network for the budding yeast, Saccharomyces cerevisiae, constructing 34 million double mutants to map ~450,000 negative and ~250,000 positive genetic interactions. The genetic interaction profile for each gene enabled quantification of functional similarity and the assembly of a hierarchical model of cellular function. Negative genetic interactions connected functionally related genes, highlighted essential genes as network hubs that provide a network scaffold for a set of core bioprocesses, and defined pleiotropic genes that mediate connections between numerous different bioprocesses. Positive interactions, especially among essential genes, appeared to reflect more general regulatory mechanisms. The organizing principles of the yeast genetic network, whereby coherent sets of positive or negative interactions occur within and between genes encoding complexes and pathways, can be exploited to decipher genetic interactions in humans.
2.20 pm - The design of the BRAIN Project Andrea Beckel –Mitchener - NIMH
Brain Research through Advancing Innovative Neurotechnologies (BRAIN) is a relatively new initiative aimed at revolutionizing our understanding of the human brain. By accelerating the development and application of innovative technologies, researchers will be able to produce a revolutionary new dynamic picture of the brain that, for the first time, shows how individual cells and complex neural circuits interact in both time and space. The NIH is working in close collaboration with other government agencies, the scientific community, as well as private partners to ensure success through investment and cooperation. One component of BRAIN is an effort to generate a census of all cell types in the brain. Significant effort is devoted to understanding what cellular properties define a given cell type, how are these properties measured and, finally, what are the challenges associated with scaling up and harmonizing large data generation efforts.
2:40 pm - GTEx: Successes and challenges Kristin Ardlie - Broad Institute
Correlations between genotype and tissue-specific gene expression levels will help identify regions of the genome that influence whether and how much a gene is expressed. The Genotype-Tissue Expression (GTEx) project aims to provide a resource to the scientific community with which to study human gene expression and regulation and its relationship to genetic variation. This project collects and analyzes multiple human tissues from donors who are also densely genotyped, to assess genetic variation within their genomes. By analyzing global RNA expression within individual tissues and treating the expression levels of genes as quantitative traits, variations in gene expression that are highly correlated with genetic variation can be identified as expression quantitative trait loci, or eQTLs. The successes and challenges of this large-scale project will be discussed.
3:00 pm - Defining scope of big projects to enable success Ajay Pillai - NHGRI
Defining scope of big projects to enable success
Ajay Pillai
NHGRI
In my presentation, I will provide an overview of considerations that go into success or failure of large-scale projects based on my experience working on multiple NIH Common Fund projects and other programs within NHGRI. Clearly articulating the goals of the project is crucial: is the project meant to answer specific questions or be a community resource? A sample of the types of questions that I will address include the following: How do you incentivize the research groups to keep overall project goals in mind throughout the period of the project? What specific incentives exist for junior scientists in each group to work collaboratively? Is the project suitable to include R01 scientists in some manner? How do you identify a community and include them in your journey during the project? What happens to the resources after the project is done?
3:20 pm Coffee Break
Session 4 Measuring Protein – Protein Interactions - Complexes and Kinetics
3:40 pm - Proteomic approaches to identify protein-protein interactions and protein complexes Ann Claude Gingras - Mount Sinai Hospital
Proteomic approaches to identify protein-protein interactions and protein complexes
Anne-Claude Gingras
Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital, Toronto, ON, Canada
We develop experimental technologies and software tools for proteomics, primarily for the analysis of protein-protein interactions. Sample quality is key to interaction proteomics success, and the choice of the expression system, epitope tag, and purification protocol can greatly influence outcome. Our lab has developed robust protocols for the purification of interactors using epitope-tagged approaches. We also implemented the BioID in vivo biotinylation approach first introduced by Kyle Roux, which enables us to probe interactions that were previously very difficult to detect (e.g. with membrane-associated proteins). Another important factor in the success of interaction proteomics experiments is the ability to discriminate true interaction partners from background contaminants. Many excellent methods exist that utilize high stringency purification procedures to limit background, or employ isotope-based quantitative proteomics approaches to assist in the identification of true interactors. However, such approaches may not always be feasible or desirable, and we perform single affinity purification coupled to mass spectrometry from unlabeled cells. Since background contaminants are generally more abundant in these cases, we are developing computational tools to allow us to identify true interaction partners. We have developed, with Alexey Nesvizhskii and Hyungwon Choi, the SAINT (Significance Analysis of INTeractome) series of algorithms that provide confidence scores for each detected interaction. The relevance of these approaches to the Human Quantitative Dynamics project will be discussed.
4:00 pm - Towards affinity quantification of protein-protein interactomes? Gilles Travé - École Supérieure de Biotechnologie de Strasbourg
Towards affinity quantification of protein-protein interactomes?
Gilles Travé
UMR 7242 CNRS, École Supérieure de Biotechnologie de Strasbourg, Illkirch, France
Most proteins establish multiple interactions with pools of alternative partners, which compete with each other based on individual binding affinities. Comparing the various affinities of a given protein to the full panel of its competing cellular partners, allows one to assess the binding specificity of this protein to each partner. Until now, high-throughput protein interactomics studies have mainly delivered binary data (“interact” or “not interact”) without quantitative information on affinities and specificities. The development of approaches for systematic determination of protein-protein interaction affinities thus represents a key challenge in systems biology. Unfortunately, most full-length proteins are difficult to produce and purify for binding assays, especially at high-throughput. Yet many protein-protein interactions boil down to interactions between globular domains and short linear sequence motifs. Most globular domains and peptides can be produced at high-throughput for interaction assays. Therefore, one way to tackle the interactome affinity determination problem is to measure domain-motif interactions at high-throughput. My talk will focus on this issue. I will in particular describe the "holdup" method for high-throughput domain-motif affinity determination, which we recently developed and applied to the PDZ domain family involved in various signaling processes (Vincentelli, Luck et al., Nature Methods 2015).
Structural and Computational Biology Unit, EMBL-Heidelberg, Heidelberg, Germany
Eukaryotic cells use membrane-bound organelles with unique lipid and protein compositions to regulate and spatially organize cellular functions and signaling. As part of this tight control, many proteins are regulated by lipids. In humans, the importance of these regulatory circuits is evident from the variety of disorders arising from altered protein–lipid interactions, which constitute attractive targets for pharmaceutical drug development. However, the full repertoire of interactions remains poorly explored and exploited because their detection is still difficult to achieve on a large and systematic scale. I will present pioneering technologies such as affinity-purification (AP)-lipidomics and the liposome microarray-based assay (LiMA) that will enable protein-lipid interactions to be deciphered on a system level. I will also illustrate the importance of these methods with two pilot studies conducted in yeast and human cell lines, which revealed surprising insights, such as the discovery of a new, conserved family of oxysterol-binding protein (OSBP) with unexpected specificities for phosphatidylserine, an important signaling lipid. The datasets also reveal cooperativity as a key mechanism for membrane recruitment of pleckstrin homology (PH) domains.
4:40 pm - Quantitative profiling of protein-protein and drug-target interactions in human cells Mikko Taipale - University of Toronto
The Donnelly Center, and Department of Molecular Genetics, University of Toronto, Toronto, ON, Canada
We are interested in systematically and quantitatively profiling protein-protein and drug-target interactions in human cells. Our focus is on the cellular quality control machinery, including major cellular chaperones, such as Hsp90 and Hsp70, and E3 ligases and deubiquitinases that collectively regulate thousands of target proteins in the cell. We are particularly interested in how these quality control pathways impinge on cellular pathways underlying rare Mendelian disorders and tumorigenesis.
5:00 pm - New technologies to study cellular quantitative dynamics Marc Birtwistle - Icahn School of Medicine at Mount Sinai
Department of Pharmacology and Systems Therapeutics and Systems Biology Center, New York, Icahn School of Medicine at Mount Sinai, New York, NY
The cellular and biochemical systems that dictate human phenotypes are large, complex and incompletely understood -- both qualitatively and more so quantitatively. Predicting human quantitative dynamics will require new experimental technologies that provide data on a much more comprehensive scale to sufficiently constrain models of such large and complex systems. This talk will identify the latest techniques that are making strides in this direction. This will include brief descriptions of the different systems-level technologies that can be scaled up to obtain comprehensive characterization of different classes of cellular components in different cell types. I will also highlight areas that need the most technology development focus.
5:20 pm - Systematizing the analysis and assembly of executable models Peter Sorger - Harvard Medical School
Systematizing the analysis and assembly of executable models
Peter Sorger
Laboratory of Systems Pharmacology, Harvard Medical School, Boston, MA
I will discuss recent approaches to the development of executable models of signal transduction (focusing on oncogenesis and drug mechanism of action) in which specialized computer languages are used to establish and maintain a hierarchy of abstractions that effectively capture both mechanistic and network-level representation of biological function. Our programs incorporate existing knowledge and are trained and tested against experimental data. Where uncertainty exists, we create families of alternative models and compare the ensemble to empirical data in a Bayesian framework that accounts for underlying uncertainties in data and assumptions; we also use executable models to optimize experimental design. The resulting reasoning system makes clear the assumptions inherent in different explanations and of necessity liberates the underlying data for re-analysis by others. Because the process from original hypothesis to experimental test to revised hypothesis is codified in open-source code and public data, findings becomes reusable, quantitative, and reproducible.
Discussion II – Experimental Technology Needs and Development
5:40 pm – 6:20 pm - Salvatore Sechi,
Discussants: Mikko Taipale, Peter Sorger, Marc Birtwistle
1. What additional experimental technologies do we need to increase throughput?
2. Can we assess what is commercially available and determine how it might fit in?
3. How do we start thinking about QA/QC issues early on? Is technology the major bottleneck, or do we have to change how we deal with QA/QC?
4. How do we identify and classify metadata associated with different experimental technologies?
6:20 pm End of the day 1
7:00 pm Reception - Cocktails and Dinner at the hotel
December 2
Session 5: Computable Databases for Human protein-protein interactome
9:00 am - Data Integration in a cell type selective manner Olga Trayanoskaya - Princeton University
Data analyses and integration in a cell type selective manner
Olga Troyanoskaya
Department of Computer Science & Lewis-Sigler Institute of Integrative Genomics, Princeton University, Princeton, NJ, and Simons Center for Data Analysis, Simons Foundation, New York, NY
An immense molecular complexity forms the foundation of human disease. This complexity must be interpreted and distilled through accurate modeling of molecular networks and pathways whose malfunction promotes the emergence of complex human diseases. Although cell-lineage-specific gene expression and function underlie the development, function, and maintenance of diverse cell types within an organism and are critical to understanding molecular basis of disease, high-throughput data are rarely resolved with respect to specific cell lineages. In this talk, I will focus on our recent work developing integrative approaches that leverage functional genomics data collections to study how cellular pathways function in diverse cell types and identify genes and systems-level changes underlying complex human disease.
9:20 am - Capturing dynamics from multiple data sets to identify rules that govern signal processing from complex input Michael J. Lee - University of Massachusetts
Capturing dynamics from multiple data sets to identify rules that govern signal processing from complex input
Michael J. Lee
Program in Systems Biology, University of Massachusetts Medical School Worcester, MA
An immense molecular complexity forms the foundation of human disease. This complexity must be interpreted and distilled through accurate modeling of molecular networks and pathways whose malfunction promotes the emergence of complex human diseases. Although cell-lineage-specific gene expression and function underlie the development, function, and maintenance of diverse cell types within an organism and are critical to understanding molecular basis of disease, high-throughput data are rarely resolved with respect to specific cell lineages. In this talk, I will focus on our recent work developing integrative approaches that leverage functional genomics data collections to study how cellular pathways function in diverse cell types and identify genes and systems-level changes underlying complex human disease.
9:40 am - Data curation and database resources to enable quantitative systems biology Ravi Iyengar - Icahn School of Medicine at Mount Sinai and Olga Troyanoskaya - Princeton University (in place of Kara Dolinski)
Data analyses and integration in a cell type selective manner
Kara Dolinski
Lewis-Sigler Institute of Integrative Genomics, Princeton University, Princeton, NJ
To produce simulations and predictions of complex physiological processes that determine human health and disease, a vast amount of data need to be generated, collected, and organized in a computable way. More specifically, a cellular parts list of proteins must be accurately and consistently annotated, along with each protein’s biochemical and kinetic properties and participation in biochemical reactions. At a higher level, these proteins need to be organized into their larger, dynamic interaction networks and pathways, along with their context (cell-type and tissue specificity). While much of these data have yet to be generated, a substantial amount of it exists, but it lies in disparate databases, supplemental data files, or embedded in the text of the biomedical literature in forms not amenable to large-scale integration and calculation. These heterogeneous data types need to be systematically curated and collated from the literature and as they are continually produced via high-throughput methods and made publicly available in a robust way to enable the predictive modeling of human health and disease that will help lead to precision medicine. Here, I will present some of the existing resources that might be leveraged toward this end, and I will also present ideas and guide a discussion on what gaps need to be filled.
Discussion III – Databases for Computing Dynamics
10:00 am – 10:45 am - Olga Troyanskaya
Discussants: Marc Vidal, Shankar Subramaniam
1. How do we resolve qualitative vs. quantitative features of databases?
2. How do we annotate common and distinct features of data for different cell types?
3. Associating metadata and QA/QC with core HQD data
10:15 am - Structured Open Discussion Olga Trayanoskaya, Marc Vidal
10:45 am Coffee Break
Session 6 Measuring Small Molecule Cellular Entities
11:00 am - One-hour proteomes and deep sequencing Joshua Coon - University of Wisconsin
Department of Chemistry, University of Wisconsin – Madison, Madison, WI
Modern biological and medicinal researchers increasingly rely on technologies for rapid and large-scale the comparative analysis of proteomes. Here, we describe the first identification of ~4,000 yeast proteins, a comprehensive yeast proteome, from just over an hour of analysis time. Next, we discuss the application of this technology to profile over one thousand proteomes and metabolomes for functional annotation of orphan genes. Prospects for extending this technology for ultrafast whole proteome quantification in human systems will also be presented. Finally, we describe technology for near complete mapping of the human proteome. With these methods we have achieved detection over 17,000 proteins in a single cell line and have detected peptides covering over 70% of each protein. Specifically, we recorded mass spectral evidence for 7 of every 10 amino acids in the human proteome. To our knowledge these data represent the deepest and most comprehensive proteomic analysis of the human proteome acquired to date.
11:20 am - Metabolomics to study enzymatic activity on a genome-wide scale Theresa Fan - University of Kentucky
Metabolomics to study enzymatic activity on a genome-wide scale
Theresa Fan
Department of Toxicology and Cancer Biology University of Kentucky College of Medicine, Lexington KY
Expertise in our laboratory integrates metabolomics with functional knockdown of key cellular components to understand quantitative relationships between cell and tissue metabolome and key reprogrammed enzymes in human cancer. This is illustrated with studies in lung cancer, which is a leading cause of cancer death in the United States and elsewhere. This disease is of particular concern in Kentucky, as Kentucky leads U.S. in both lung cancer incidence and fatality. The past decade of research in cancer metabolism reveals the value of exploring human metabolome for the discovery of novel therapeutic and diagnostic biomarkers for human cancers and other diseases. Using 13C6 glucose as tracer and NMR and MS-based stable isotope-resolved metabolomics (SIRM) analysis, we defined metabolic reprogramming in lung tumor tissues in 59 human patients with early stage NSCLC in vivo. In particular, we uncovered elevated anaplerotic pyruvate carboxylase (PC) activity in cancerous against paired benign tissues. Proliferating cancer cells require an active Krebs cycle for generating anabolic precursors, in addition to energy production. Diversion of the Krebs cycle intermediates to meet anabolic demands cannot be sustained without anaplerosis. Pyruvate carboxylation is one of the two major anaplerotic pathways that support cancer cell growth; the other involves glutaminolysis initiated by glutaminase (GLS). We then measured the expression of PC and GLS proteins in tumor and benign tissue pairs from 86 human NSCLC patients. PC protein was elevated (median 8-10 fold) in 94% of the tumor tissues, whereas GLS expression did not differ significantly between tumor and benign tissue pairs. We further examined the importance of PC in the growth and survival of NSCLC by employing the shRNA knockdown approach to demonstrate reduced lung cancer cell growth both in vitro and in vivo. PC knockdown not only attenuated PC-initiated Krebs cycle activity but also glutamine metabolism, as probed by 13C6-glucose and 13C5, 15N2-glutamine tracers coupled with SIRM analysis. These effects, in turn, led to reduced anabolic activity, such as the synthesis of fatty acyl chains of lipids, suggesting that both energy production and anabolic pathways were hindered, and blocking the PC pathway was not compensated by GLS activity. Applying the SIRM approach to ex vivo lung tissue slice cultures indicated that blocking PC anaplerosis with anti-cancer Se compounds can elicit massive necrosis with reduced mitotic index in tumor but not in paired benign lung tissues. Together, these results suggest that PC can be a promising therapeutic target for lung cancer. Such approaches can be very valuable in uncovering and deciphering the function of dysregulated cellular components in cells or directly in individual human tissues on a metabolome and genome-wide basis.
11:40 am - Quantitative profiling of lipid metabolities Robert Murphy - University of Colorado
Department of Pharmacology, University of Colorado School of Medicine, Aurora, CO
Lipids constitute a very large family of biomolecules with hydrophilic as well as hydrophobic properties that can be divided into six different groupings found in mammalian systems. These include fatty acyl compounds, glycerolipids, glycerophospholipids, sphingolipids, steroids, and prenol lipids. These molecules play critical roles in mammalian biochemistry, not only as sources of energy and ATP through fatty acid oxidation, but also physical-chemical roles as the major constituents of membranes that compartmentalize cells as well as subcellular vesicles. Also, some lipids are bioactive mediators, including prostaglandins, thromboxanes, leukotrienes, ceramides, phospholipids, and steroids. These diverse roles as well as the complexity with which lipids present themselves in the living cell requires demanding analytical techniques to both qualitatively and quantitatively assess their levels. Mass spectrometry has emerged as the predominate analytical technique largely because of the development of electrospray ionization, which makes analysis of all lipid substances possible. It is also possible to image the location of lipids within tissues and identify them with a high degree of structural specificity. The field of lipidomics has emerged as a powerful tool to understand the complexity of lipid biochemistry, the potential role of bioactive lipid mediators in physiology as well as pathophysiology. In this presentation, I will present approaches to comprehensively and quantitatively profile lipids in different human cell types.
12:30 pm Lunch
Session 7 Integrating Informatics with Modeling Tools
1:40 pm - Integrative Informatics for cellular metabolites and macromolecules Shankar Subramaniam - University of California
Integrative Informatics for cellular metabolites and macromolecules
Shankar Subramaniam
Department of Bioengineering, University of California, San Diego, San Diego, CA
My laboratory has extensive experience in developing and organizing large data sets and developing informatics tools. The Biology Workbench (the first web-based integrative infrastructure for protein and nucleic acid sequence analysis), the Signaling Gateway (the repository of functional states of signaling proteins), the Lipidomics Gateway (the most comprehensive resource for lipids), the Metabolomics Workbench (the NIH Common Fund’s metabolomics resource for the scientific community), and more recently, the Alzheimer’s Workbench (an integrative infrastructure for multiscale data and analysis for Alzheimer’s disease) are examples of infrastructures developed in my laboratory. The metabolome can be regarded as end-point readout of cellular function with RNA and protein components interplaying to regulate, synthesize, and degrade metabolites. In essence, the metabolome by itself, or in conjunction with the transcriptome and proteome, can define the state of a physiological system and longitudinal measurements of these components provide insights into the dynamical changes in the state of the system. Additionally, the availability of absolute standards for thousands of metabolites provides a quantitative framework for measuring changes in metabolites, and this enables a kinetic analysis of a cellular phenotype. In this talk, I will address the challenges in developing infrastructures for cellular components focusing mainly on metabolites, discuss the challenges in quantitative measurements, and finally describe case studies where the integrative analysis of multi-omics data provides dynamical models of cellular function.
2:00 pm - Modeling approaches for whole cell dynamics: Challenges and opportunities Chad Myers - University of Minnesota
Modeling approaches for whole cell dynamics: Challenges and opportunities
Chad Myers
Department of Computer Science and Engineering, University of Minnesota, Minneapolis, MN
One of the goals of systems biology is to build comprehensive, mechanistic models of the cell. After more than a decade of collecting genome-wide data capturing gene expression, protein levels, and physical or genetic interactions among proteins/genes, we have accumulated a wealth of data that should help us make progress towards this goal. As a community, we have made great strides in developing computational methods that leverage these genomic data to accomplish various prediction tasks, e.g. gene function or protein expression/localization prediction. While all of these are of great utility, I would argue that we are still quite far from systematic approaches for building mechanistic models that, for example, allow us to explain or simulate quantitative, dynamic behavior of biological systems. In my opinion, this highlights the fundamental difficulty of the mechanistic modeling problem, both in terms of the computational approaches required as well as in the focused, quantitative data necessary for the inference of such models. From the perspective of a few specific projects we have been involved in, I will highlight a few of the challenges we have encountered in making the transition between statistical and mechanistic models, and discuss opportunities for new research.
2:00 pm - Modeling and analysis tools for whole cell modeling Les Loew - University of Connecticut
Modeling and analysis tools for whole cell modeling
Les Loew
Departments of Cell Biology, Computer Science and Engineering, University of Connecticut, School of Medicine, Farmington, CT
The shape of a cell, the size of subcellular compartments, and the spatial distribution of molecules within the cytoplasm can all control how molecules interact to produce a cellular behavior. This talk describes how these spatial features can be included in mechanistic mathematical models of cell biology. The Virtual Cell (VCell) computational modeling and simulation software is designed for spatial modeling of cellular reaction-diffusion systems. VCell facilitates choices between physical formulations that implicitly or explicitly account for cell geometry and between deterministic vs. stochastic formulations for biochemical reactions. It is also built on top of a database to facilitate sharing of models and reuse of model components. Recently, we have begun to address the special role of molecular aggregates and clusters in cell biology. New approaches for dealing with the combinatorial complexity associated with interactions of multivalent molecules will be described, including studies on signaling to the actin cytoskeleton.
Discussion IV: Measurements for Inferring Dynamics/ Integration of Data Gathering and Informatics with Modeling
Program in Systems Biology, University of Massachusetts Medical School, Worcester, MA
The nematode C. elegans is a powerful model for the systems-level study of development and homeostasis. Its power comes from its short generation time, transparent body, outstanding genomic and genetic resources, and low cost of experimentation, among others. I will discuss how a pilot project of quantitative dynamics can be constructed for the worm, with a main focus on the integration of metabolic, transcriptomic and interactomic information, and how this relates to anabolic processes during animal growth and fecundity and catabolic processes to generate energy to support cellular and organismal activities. We propose that such a C. elegans project will demonstrate the power of a quantitative dynamics program, and that it will provide a blueprint to guide similar studies in more complex systems.
3:20 pm - Multiscale modeling: Understanding adaptive immune responses in terms of dynamic molecular networks within each cell Alexander Hoffmann - UCLA
Multiscale modeling: Understanding adaptive immune responses in terms of dynamic molecular networks within each cell
Alexander Hoffmann
Institute for Quantitative and Computational Biosciences (QCBio), UCLA, Los Angeles, CA
Over the past decades, kinetic models have provided important quantitative insights in immunology at two scales: (1) kinetic models of lymphocyte populations have led to insights about the critical interplay between different lymphocyte subsets during chronic and acute infections; and (2) they helped reveal how a wealth of regulatory dynamics are generated in the molecular networks that determine lymphocyte function. In our recent work, we have begun to link these two scales through multiscale modeling that allow us to probe how the dynamics of molecular networks in each cell determine cellular cell fate decisions, and how these contribute to population-level functions of immune responses. I will outline how determination of kinetic parameters of key molecular networks would enable the development of a multiscale model of human immune system that will mechanistically link molecular and physiological scales, and thus, down the line, provide predictive power to interpret clinical presentations in the context of genetic variants and exposure histories, identify drug targets, and inform clinical decision making.
3:40 pm - A human iPSC-derived cardiomyocytes cell project - all it takes to make & keep a beat Ravi Iyengar - Icahn School of Medicine at Mount Sinai
A human iPSC-derived cardiomyocytes cell project - all it takes to make & keep a beat
Ravi Iyengar
Cardiomyocytes in culture can form functional sarcomeres and spontaneously beat, thus representing a cell type where there is a clearly observable phenotype to which cellular components and interactions can be correlated to a physiological function. Induced pluripotent stem cells (iPSCs) from human subjects can be readily differentiated into cardiomyocytes that beat. As the iPSCs can be derived from individual human subjects, the technology currently exists to obtain functionally active cardiomyocytes with well-characterized genomes. As part of the LINCS program, our DToxS Center has created a library of iPSC lines from 20 clinically-characterized healthy human subjects. These cell lines can be grown up in a large-scale manner, and they could potentially serve as the source material for a HQD pilot project. The iPSC-derived cardiomyocytes are primary cells, that once differentiated, do not divide; hence, they could provide a useful contrast to cell lines that divide in culture and thus may have an underlying proliferative program. Among the considerations in using iPSC-derived cardiomyocytes is the level of their differentiation: whether they adequately represent adult cells or more like neonatal or fetal cardiomyocytes. Both the pros and cons of considering iPSC-derived cardiomyocytes for a pilot project will be discussed.
Discussion V: Conclusions and the Way Forward
4:00 pm – 5:00 pm Sarah Dunsmore & Marian Walhout
Discussion: Olga Troyanskaya, Alex Hoffmann, Ravi Iyengar, Ajay Pillai, Salvator Sechi, David Balshaw
1. Reproducibility and applicability in differing contexts: Structuring large-scale projects for success, defining and constraining scope
2. Community engagement and how big projects can enable small projects
3. Path forward engaging various interested parties