Joel Graber
Genomic Domain and Network Characterization
Multiple lines of evidence point to a distinctly non-random organization of functional elements in the genome. It appears that the mouse genome is organized into domains-contiguous, possibly overlapping, segments of chromosomes containing functionally related groups of elements (including, but not limited to genes and regulatory elements) and that non-contiguous domains and isolated elements interact in networks. This project is developing the bioinformatics tools required for using multiple, diverse data sets in analyzing these relationships.
We are:
- Developing methods and software to identify and characterize domains, moving from analyses based on dynamic programming to more sophisticated methods based on Hidden Markov Models. We are including methods for combining multiple data sources to form composite domain structures.
- Generating the networks implied by disparate data, using a graph representation in which nodes are chromosome positions (or intervals) and edges imply an interaction (direct or indirect) between the nodes. As with domains, we will generate both data-specific and composite networks for analysis. We are developing methods and software to compare alternative networks, focusing specifically on methods to identify statistically significant common subnets.
- Providing our computational tools to the other projects in the Center for the analysis of data generated by the Center [e.g., linkage disequilibrium data from SNPs, gene expression data, recombination data, as well as overall integration and data obtained from external sources (e.g., GO annotations or the KEGG database). We are working also with the Center’s Computational Core to define efficient and comprehensive databases and web interfaces for the data and analysis we produce.
