Let's take a look at the
process by
which one produces data using a microarray. We start with two kinds of
genetic source material.
First, we acquire clones each of which produces a single specific
cDNA sequence that is known in advance. These are reproduced in large
quantities and spotted onto an array in a regular geometric pattern by
a robot. The array can be made on a glass slide or a nylon
filter. Each spot of cDNA is called a probe.
Second, we acquire mRNA from a target sample of
interest. This sample must be labelled either radioactively or with a
fluorescent dye.
The basic experiment is then to hybridize the target RNA with the
DNA probes. Complementary sequences will stick together. After some
time passes, any excess RNA will be washed off.
Up to this point, we've been describing biology. (You can tell,
because everything involves wet, messy, sloppy stuff.) Now, however,
we've converted everything into nice, clean computer files, and
entered the realm of informatics. The first informatics problem is to
quantify the spots. This typically involves aligning an abstract
template to the image and using your favorite algorithm to wiggle it
about until the "ideal spots" match up reasonably closely with the
"actual spots" on the image. Numerous measurements can be made at each
spot, and they need to be stored somewhere (preferably in a
database where you can find them later).
Next, we usually need to normalize the raw data. That may just
mean performing background correction, but it may be more elaborate:
different experimental conditions, such as time of hybridization opr
time of exposure to the phosphorimager, can lead to differnce in image
intensity that have nothing to do with iunderlying biology, and the
data may need to be adjusted to remove such systematic errors.
Finally, you're ready to start using statistical and mathematical
methods to analyze the data.
Disclaimer:
This page was last updated on 12 October 2002. It is entirely possible
that the information contained herein no longer has any connection
with reality (assuming it ever did). Feel free to send
constructive comments or
inane criticisms to:
Kevin Coombes
Departments of Biostatistics and Biomathematics
University of Texas M. D. Anderson Cancer Center
Houston, TX 77030