Network Inference Testbed
Ron Taylor, Principal Investigator
There are currently no interactive environments available to (1) evaluate different algorithms for inference of biological regulatory network structure using common data sets and (2) easily apply such state-of-the-art algorithms to experimentally generated, high-throughput data. The goal of the Network Inference Testbed (NIT) project is to fill this gap. The NIT research team at Pacific Northwest National Laboratory (PNNL) is constructing a software platform that permits direct inference of:
- genetic regulatory networks from high-throughput microarray, messenger RNA (mRNA) expression data
- protein regulatory networks from high-throughput, protein expression data.
Methods to analyze high-throughput data search for patterns of partial correlation or conditional probabilities that indicate causal (regulatory) influence in the input set of expression values. Such patterns of partial correlations found in the high-throughput data, possibly combined with other supplemental data on the organism of interest, are the bases upon which the algorithms in the NIT's toolkit will infer regulatory networks.
Within the NIT software environment, the user can test the performance of the different inference algorithms on artificial gene and protein expression data sets, which contain simulated perturbations. A suite of optimized and trained inference algorithms, or toolkit, can then be identified and used to analyze experimental, high-throughput expression data. This toolkit of inference methods will be deployed within a common framework; an application program interface (API) will allow the user to add other inference algorithms to the toolkit.
This research will create a suite of software tools, which will access a common relational database and enable computational biologists to rapidly reconstruct biological regulatory networks with greater ease and accuracy. Hence, the testbed will be useful both to software developers wishing to compare, refine, or combine inference techniques and to bioinformaticists analyzing experimental data.
