Biology Reference
In-Depth Information
to extend the utility of clique-centric methods that may produce them. Using
non-obese diabetic mice as a target organism, the paraclique algorithm is tested
on transcriptomic data under various parameters in order to determine how it can
best be tuned to applications. The use of proteomic anchors is also discussed
in an effort to help guide subgraph selection in the presence of inhomogeneous
data, which is an important but notoriously difficult problem in its own right.
10.1. Overview
Many inbred strains of
Mus musculus
, the common house mouse, are employed
in biomedical research. The non-obese diabetic (NOD) mouse is particularly use-
ful as a model of type 1 diabetes mellitus (also called juvenile onset, or insulin
dependent, diabetes). In both mice and humans, this disease is characterized by
persistent hyperglycemia (elevated blood sugar level) that is induced in geneti-
cally susceptible individuals and modified by a variety of environmental triggers
including food and infections. It is caused by an abnormal and self-destructive
immune response (autoimmunity), which allows mononuclear leukocytes to tar-
get the insulin producing beta cells in the pancreas [26, 27, 32]. Eventually this
process destroys so many of the beta cells that the body is unable to produce suf-
ficient insulin to retain normal blood glucose levels and diabetes is observed. Our
studies in the NOD mouse focus on the very early leukocyte abnormalities that
may be associated with initiation of the autoimmune process [14]. If we can gain
a better understanding of the initiation of autoimmunity, then we may be able
to develop rational intervention strategies that can stop the disease process in its
preclinical phase effectively and with minimal side effects.
The importance of melding experimental research with continuing advances in
computational analysis is well understood [17, 18, 21]. In the work reported here,
we begin with high-throughput NOD mouse data and apply novel clique-centric
methods to analyze it. Fixed parameter tractability [1, 8] and various realizations
of the paraclique algorithm [6] form the basis of techniques we use to extract
dense putative networks from the vast sea of correlations that arise in the anal-
ysis of comprehensive transcriptomic data [4, 19]. Proteomics data is added to
the mix, thereby introducing challenging new problems in inhomogeneous data
interpretation [16]. The results we obtain are evaluated in terms of both statistical
quality and biological relevance.
Search WWH ::
Custom Search