The Database of Quantitative Cellular Signaling: management and analysis of chemical kinetic models of signaling networks

BIOINFORMATICS Vol. 19 no. 3 2003, pages 408–415 DOI: 10.1093/bioinformatics/btf860 The Database of Quantitative Cellular Signaling: management and analysis of chemical kinetic models of signaling networks Sudhir Sivakumaran, Sridhar Hariharaputran, Jyoti Mishra and Upinder S. Bhalla ∗ National Centre for Biological Sciences, GKVK Campus, Bangalore 560065, India Received on February 15, 2002; revised on June 28, 2002; accepted on September 11, 2002 INTRODUCTION Signaling networks are the computational and control system of the cell. The traditional view of signaling involves transduction of chemical signals at the cell surface, and their propagation via sequences of biochemical events involving proteins and second messengers (Stryer, 2001). In its broader sense, the cellular signaling network includes ∗ To whom correspondence should be addressed. 408 genetic, cytoskeletal and cell trafficking elements. Qualitative analyses of such networks have been carried out using logical representations, for example in plant signaling (Mendoza et al., 1999; reviewed in Genoud et al., 2001). These studies are appropriate in many cases where general connectivity is understood, but kinetic details are uncertain. Mathematical and modeling methods are useful in gaining further insights into cellular function (Tyson et al., 2001). Several recent studies have undertaken a quantitative analysis of cellular signaling at the level of massaction kinetics of signaling pathways (Bhalla and Iyengar, 1999; Kuroda et al., 2001; Lamb, 1994) and genetic interactions (Gillespie, 1977). Some studies include analysis of the three-dimensional, stochastic and cellular mechanical function (Stiles et al., 1998; Shimizu et al., 2000; Arkin and Ross, 1994). Current data sources, such as testtube biochemistry, are clearly poor approximations to cellular conditions, nevertheless these are the best sources of data we currently have. It is anticipated that more biologically detailed descriptions will become increasingly feasible with new experimental techniques (Teruel and Meyer, 2001; Kierzek, 2001; Voytik-Harbin et al., 2001). All such quantitative descriptions have mass-action chemistry as a common denominator. There is therefore a clear need for developing data management and analysis systems appropriate for such data. A number of initiatives have come into being as part of this process. They can broadly be grouped into three categories: databases, simulators and model description languages. Although it is structured as a database, the DOQCS project draws its inspiration especially from simulation projects related to cellular signaling (e.g. Bhalla and Iyengar, 1999). These studies have given rise to many models incorporating detailed and explicit reaction schemes and parameters. DOQCS is a resource for such models that sets quantitative functional analysis of the data as a central consideration for database development. As this emphasis differs from other databases, a further goal of the DOQCS project was to identify and fulfill c Oxford University Press 2003; all rights reserved. Bioinformatics 19(3)  ABSTRACT Motivation: Analysis of cellular signaling interactions is expected to pose an enormous informatics challenge, perhaps even larger than analyzing the genome. The complex networks arising from signaling processes are traditionally represented as block diagrams. A key step in the evolution toward a more quantitative understanding of signaling is to explicitly specify the kinetics of all chemical reaction steps in a pathway. Technical advances in proteomics and high-throughput protein interaction assays promise a flood of such quantitative data. While annotations, molecular information and pathway connectivity have been compiled in several databases, and there are several proposals for general cell model description languages, there is currently little experience with databases of chemical kinetics and reaction level models of signaling networks. Results: The Database of Quantitative Cellular Signaling is a repository of models of signaling pathways. It is intended both to serve the growing field of chemicalreaction level simulation of signaling networks, and to anticipate issues in large-scale data management for signaling chemistry. Availability: The Database of Quantitative Cellular Signaling is available at http://doqcs.ncbs.res.in. Links to the signaling model simulator, GENESIS/Kinetikit are at http:// www.ncbs.res.in/∼bhalla/kkit/index.html and are also provided from within the database. The database source code is available under the GNU Public License. Contact: Database of quantitative cellular signaling distinctive database requirements as the dataset expanded and usage patterns became clearer. DESIGN AND IMPLEMENTATION Data model Chemical kinetic simulations are performed by converting chemical equations of the general form kf A + B
 C + D kb to systems of differential equations of the form d[A]/dt = −k f [A][B] + kb [C][D] and applying standard numerical integration methods (Bhalla, 1998) to calculate the time evolution of these reactions. There are several ways of specifying chemical kinetic models. Many models are reported in terms of concise systems of differential equations, after applying mass conservation rules to eliminate redundant equations (e.g. Asthagiri and Lauffenburger, 2000). It is also common to assume equilibrium relationships between molecules to avoid solving multiple differential equations (Grzybowski et al., 2000; Hecht et al., 1990). Some models have also been described in terms of concentration-dependent rate constants (Kholodenko, 2000). In distinction to these abstractions and numerical simplifications, simulators such as GENESIS/Kinetikit take a more chemically detailed approach and require that every molecule, reaction, and enzyme activity be explicitly specified (Bhalla, 2002a). A key design decisions for DOQCS was to retain this explicit chemical-level description for • There is a direct correspondence between database entries and experimentally measurable quantities such as reactant concentrations. • As there are no assumptions about equilibrium situations it is possible to apply reaction schemes to dynamic chemical situations on a time-scale shorter than the equilibrium time-scale. • Stochastic chemical systems can be represented without any change to the database. It is sufficient to re-interpret the entries for kinetic rates as probabilities of reaction events. For example, a rate constant k f in units of µM/sec could be scaled to # of molecules/sec and this can be used to estimate reaction transition probabilities. The reaction scheme and other entries in the database would be unaffected. • There is a logical extension into three-dimensional reaction-diffusion systems by addition of spatial distribution information. The chemical organization of the tables need not change. DATABASE STRUCTURE The table structure of the database is designed t (...truncated)