MICROARRAY DATA ANALYSIS IN AFFYMETRIX GENE CHIP - Quantitative Medical Data Analysis Using Mathematical Tools and Statistical Techniques - page 147

2.3.4. Classication by Keywords
Classication based on GO annotations presents useful information to study
gene association. However, the lists of annotations may remain large. In-
vestigators would want to narrow down the lists and focus on specic areas
of interest. In addition, GO tries to synchronize the description of genes,
but there are still wide variations in genomic terminology. Gene alias is an
obvious case. For example, the secreted phosphoprotein 1 (SPP1) is a gene
associated with ossication. It has various names, such as OP, Bsp, Eta,
Ric, Apl-1, and minopontin. These names are quite textually dierent from
the ocial name. Without the alias information, it would be dicult to
relate the various alias names to this gene if they are used. To eectively
correlate genes with keywords, we consider various variables, in addition to
GO annotations, to search genes associated with the keywords. The cur-
rent variables included in the search-database are GO annotations, gene
alias, gene name, gene description, KEGG pathway, and RIF. Given the
keywords, the tool will search the database and identify the genes associ-
ated with these keywords. A summarized table of keyword classication will
be given. Each keyword classication will include a set of genes associated
with the keyword. The generated table will help investigators expedite the
process of gene function classication.
2.3.5. Implementation
The approach has been written in R software 68 (available on request by
email at [e-mail]). The outputs are a series of html les. Starting
with the main.htm le at the root directory, it will guide the user to the
whole set of regulated genes, GO classication, or keyword classication.
3. Data Example: A Prolactin Study
3.1. Background and Study Design
Prolactin (PRL) is a lactotrophic hormone synthesized and secreted by the
pituitary gland 69 . It has been indicated to be associated with regulation
of the outgrowth of new capillary blood vessels, a process referred to as
angiogenesis 70 . In particular, 23 kDa prolactin (23k PRL) has been shown
to be angiogenic while its proteolytic fragment 16 kDa prolactin (16k PRL)
has been shown to be antiangiogenic 71 . To address how signaling by 23k
and 16k PRL aects endothelial cell function, gene array analysis was per-
formed in the following settings. Recombinant adenovirus expressing the