Abstracts
Patricia Babbitt (QB3/UCSF, San Francisco) "Network Analysis Reveals New Relationships among Divergent Proteins Missed by Tree-based Approaches" Availability of large-scale data from the genome projects has inspired development of sophisticated new methods for inference of molecular function using sequence and structure-based comparisons. Management and analysis of very large sets of distantly related protein sequences has become increasingly difficult with currently available tools, however, as the most robust methods for generating multiple alignments or phylogenetic reconstruction cannot generally handle thousands of sequences. Further, the difficulty of obtaining a sufficiently good alignment for phylogenetic reconstruction of highly divergent sequences or structures frequently diminishes the value of the trees that can be obtained. We present preliminary results using Cytoscape as an alternate approach for clustering and analyzing such data and show that even using simple distance metrics such as Blast E values provides useful insights about sequence and functional links among large sets of remotely related proteins. Moreover, network analysis can reveal multiple connections among clusters of related sequences that cannot be accessed from multiple alignments or trees.
Gary Bader (University of Toronto) "BioPAX, cPath, and PDZ-based Protein Interaction Prediction" Cellular biological systems, such as cell signaling networks which mediate cellular information transfer and logic, are implemented by precisely controlled specific molecular interactions. For example protein post-translational modifications or change of interacting molecule location can turn on or off an interaction. Evolutionary genome sequence changes or the presence of certain SNPs can also create or destroy molecular interactions. Can we use information about how these interactions are created and controlled to accurately predict biologically relevant network links? Can we find out how interactions are rewired to cause disease? To answer these questions, we require large amounts of cell map information (networks and pathways), algorithms and tools for visualization and analysis. I will discuss our efforts to develop software systems for cellular biological network analysis and the application of some of these tools to predict protein-protein interactions involving human PDZ domains.
Mirit Aladjem (National Cancer Inst., NIH) "Molecular Interaction Maps of Biological Signalling Networks" Understanding the functional mechanisms underlying complex bioregulatory pathways is essential if we wish to utilize modern molecular biology most effectively against diseases, such as cancers, that arise from defects in cell regulation. A Molecular Interaction Map (MIM) is a diagram convention capable of unambiguous representation of networks containing features that are challenging to represent, such as multi-protein complexes, protein modifications, and enzymes that are substrates of other enzymes. This graphical representation makes it possible to view all of the many interactions in which a given molecule may be involved, and it can portray competing interactions, which are common in bioregulatory networks. Interactive electronic molecular interaction maps (eMIMs) allow users to navigate through the molecular interaction network and link to molecular databases, references and annotations that contain pertinent information. Representing complex molecular interactions in detail may help discern the common patterns of molecular interaction logic that convey bioregulatory networks their remarkable flexibility and robustness.
Ilya Shmulevich (ISB, Seattle) "Data Integration and Genetic Network Modeling"
Benno Schwikowski (Institute Pasteur, Paris) "Mass Spec Informatics: Towards Dense Protein Expression Networks" Global measurements of protein abundance will be a critical ingredient for dynamic models of molecular interaction networks. Liquid chromatography, coupled to mass spectrometry (LC-MS), represents the current platform of choice for such global measurements. In single experiments, the minimum time required to identify single proteins, and the maximum time available for one experiment limit the number of identified proteins in a single experiment. This number is estimated to be too low by 2-3 orders of magnitude if one wants to be able to cover the human proteome. Here, we present an alternative approach that relies on linking information across independent experiments to identify and quantify peptides. The approach relies on the alignment of multiple LC-MS experiments. We show basic ideas, experimental validation, and a side product that systematically uncovers quality issues with this type of experiments.
Trey Ideker (UCSD, San Diego) "Protein Network Comparative Genomics" With the appearance of large networks of protein-protein and protein-DNA interactions as a new type of biological measurement, methods are needed for constructing cellular pathway models using interaction data as the central framework. The key idea is that, by comparing the molecular interaction network with other biological data sets, it will be possible to organize the network into modules representing the repertoire of distinct functional processes in the cell. Three distinct types of network comparisons will be discussed, including those to identify:
- Protein interaction networks that are conserved across species
- Networks in control of gene expression changes
- Networks correlating with systematic phenotypes and synthetic lethals
Using these computational modeling and query tools, we are constructing network models to explain the physiological response of yeast to DNA damaging agents.