Biology Reference
In-Depth Information
threshold (transversal lines). In the module layer presentation
[Fig. 8(b)], modules identified at the same resolution are represented on
a plane such that their distances reflect the regulatory relations. Modules
induced under similar conditions are closer to each other, while large dis-
tances indicate inverse activation.
4. Outlook
High-throughput data acquisition technologies have created the
potential for new insights into biological systems. Yet, the hope to bet-
ter understand the regulation of these systems and to eventually pre-
dict their response will only materialize with adequate computational
tools to process and visualize the vast amount of data that these tech-
nologies produce. Medical research is rapidly adopting high-through-
put technologies to characterize clinical samples, and therefore
requires appropriate computational tools to interpret the results and
use them for diagnostic purposes; we believe that a modular approach
to large-scale data will yield great benefits with respect to both aspects.
In particular, we anticipate a great need for integrating data from mul-
tiple platforms ranging from SNP chips and CGH chips to mRNA
chips and protein chips. Extending the modular approach for the
analysis of multiple datasets of various types has been shown to be use-
ful for predicting drug-gene associations from relatively inexpensive
high-throughput data, as generated for the NCI60 panel.
56
Such an
approach enables large-scale hypothesis generation that may allow
more cost-effective targeting of research resources for direct experi-
mental studies.
New modular approaches for integrative analysis of large-scale data
are likely to be useful also for the meta-analyses of other large-scale
biological data. For example, comodule analysis has great potential for
the integration of expression data from two different species. In this
case, the common dimension of the two sets of expression data is estab-
lished by the orthologous genes (rather than the samples, as for the
NCI-60 data). Here, comodule analysis could provide a sensitive
means to identify not only the sets of orthologs with conserved coex-
pression, but also the respective experimental conditions under which
