www.ncbi.nlm.nih.gov/pubmed/21596793
ABSTRACT
Motivation: Recent abundance of data from studies employing
high-throughput technologies to reveal alterations in human disease
on genomic, transcriptomic, proteomic, and other levels, offer the
possibility to integrate this information into a comprehensive picture
of molecular events occurring in human disease. Diversity of data
originating from these studies presents a methodological obstacle in
the integration process, also due to difficulties in choosing the opti-
mal unified denominator that would allow inclusion of variables from
various types of studies. We present a novel approach for integra-
tion of such multi-origin data based on positions of genetic altera-
tions occurring in human diseases. Parkinson disease (PD) was
chosen as a model for evaluation of our methodology.
Methods: Datasets from various types of studies in PD (linkage,
genome-wide association, transcriptomic and proteomic studies)
were obtained from online repositories or were extracted from avail-
able research papers. Subsequently, human genome assembly was
subdivided into 10kb regions, and significant signals from aforemen-
tioned studies were arranged into their corresponding regions ac-
cording to their genomic position. For each region rank product
values were calculated and significance values were estimated by
permuting the original dataset.
Results: Altogether, 179 regions (representing 33 contiguous ge-
nomic regions) had significant accumulation of signals when p-value
cut-off was set at 0.0001. Identified regions with significant accumu-
lation of signals contained 29 plausible candidate genes for PD. In
conclusion, we present a novel approach for identification of candi-
date regions and genes for various human disorders, based on the
positional integration of data across various types of omic studies.
No comments:
Post a Comment