Conferences & Workshops
The Network Assembly and Mathematical Modeling Core provides a world-class suite of programs dedicated to bioinformatic research, service, and training within the San Diego Center for Systems Biology. Its mission is to develop and apply technology for integrating diverse ‘omics data sets into maps of biological networks and to establish the means by which this information can be translated to formulate general biological models capable of accurate predictions of cell states and phenotypes. This mission is accomplished through collaboration and service with the four research projects; development of timely new services through cutting-edge bioinformatic research; a portfolio of training activities; and software and hardware computing platforms.
The core is directed by Drs. Trey Ideker and Lev Tsimring. Dr. Ideker, a Professor at UCSD in the Department of Medicine, will oversee core research and direct activities related to networks, ontologies, and associated predictive modeling. Dr. Ideker’s career has focused on systematic methods to elucidate molecular networks, and for using knowledge of these networks to guide the development of novel diagnostics and interventions. Dr. Tsimring, a Research Scientist at UCSD and the Associate Director of the BioCircuits Institute, will direct activities related to deterministic and stochastic mathematical modeling of kinetic processes in networks and integration of ‘omics data in development and validation of mathematical models of biological systems. Dr. Tsimring is a leader in the field of quantitative biology where he brings his expertise in nonlinear dynamics and biological physics to advance understanding of biological processes.
Biomedical research is increasingly dependent on knowledge of biological networks of multiple types and scales, including gene, protein and drug interactions, cell-cell and cell-host communication, and vast social networks. This service will assist projects with integration of data into unified biological network maps. We will generally follow the prevailing approach to build integrated interaction networks, which uses a probabilistic framework to compute the likelihood that a functional relationship exists between a pair of genes given the available evidence. Current efforts in this arena are exemplified by STRING, GeneMania (Morris Lab, U. Toronto), MAGIC (Troyanskaya Lab, Princeton), and HumanNet (Marcotte Lab, U. Texas). The various lines of evidence are weighted to best recover a gold-standard training set of gene pairs that function in the same pathway (e.g. KEGG or Reactome DBs).
The core will assist SDCSB investigators with a range of techniques that have been developed over the past decade for specific tasks in network analysis, including gene function prediction, detection of protein complexes and other modular structures, network evolutionary comparison, identification of active subnetworks, subnetwork-based diagnosis and subnetwork-based gene association.
It is well known that molecular networks are not merely modular (i.e., a list of complexes) but also hierarchical and multiscale; protein complexes are subunits of larger complexes, which in turn nest within pathways and organelles. We have recently found that molecular interaction networks contain rich hierarchical structure, which can recapitulate much of the hierarchy in the manually curated Gene Ontology (GO) (Dutkowski et al., Nat Biotechnol 2013). To construct a data-driven ontology, molecular network maps will be clustered using a hierarchical probabilistic model for community detection. Gene sets, represented by nodes in the tree, are suggestive of biological entities or ‘terms’ in an ontology. The resulting directed acyclic graph is interpreted as the NetworkeXtracted Ontology (NeXO). We will compare NeXO to literature-curated ontologies such as GO using ontology alignment methods such as ASMOV. Alignment to a reference ontology allows for transfer of definitions from GO to NeXO, identification of novel terms not in GO, and reveals consistent and conflicting term-term relations.
There are several distinguishing features of the approach to mathematical modeling taken by the Core. We are particularly interested in studies of dynamical behavior of the underlying models – whether deterministic or stochastic – since temporal variability and adaptation are essential properties of systems responding to transient changes in the environment. We are also interested in using a wide range of ‘omics data to build models across multiple levels of resolution: from individual genes and proteins, to protein complexes and signaling pathways, to whole systems. Finally, the core has extensive expertise in quantitative analysis of stochastic kinetic models and in performing information theoretic analyses of signal transduction in fluctuating and uncertain environments.
Mathematical or statistical models often have many parameters, some of which are known from experimental measurements or literature reports while others are unknown. Fitting the behavior of models to data helps identify unknown parameters, or at least identify the ranges in which they can vary, while preserving the desired or observed network function. Our experience in modeling synthetic gene oscillators demonstrates that matching model predictions with observed dynamic data is more constrained for parameter identification than matching homeostatic conditions (fixed points). The Core will provide SDCSB researchers with a variety of tools such as standard classification and regression analyses, stability and bifurcation analyses, robustness and sensitivity analyses, and formulations that optimize its performance in the absence of complete a priori knowledge of underlying parameters. The end result is a deterministic or stochastic dynamical model capable of generating predictions for the system’s function in novel stationary or dynamic conditions (e.g., nutrient limitation or other types of stress, genetic manipulation, interaction with other systems). When possible, these predictions will be tested against high-throughput data analysis, and, if necessary, will be used for model improvements.Submit a request for services here