Biology Reference
In-Depth Information
protein atlas for all proteins expressed in the human genome to provide
information toward the tissue-specific expression of the proteins in health
and disease (
Berglund et al., 2008; Ponten et al., 2008; Uhlen et al., 2005,
2010
). To do this, tissue microarrays representing 48 human tissues and 20 dif-
ferent cancer tumor types have been constructed and are being labeled via use
of antibodies produced, internally, by the Human Protein Atlas research
group, or antibodies obtained from other sources, such as commercial ven-
dors. There are typically three different antibodies tested for each protein,
and images from each test tissue and antibody (visualized with bright field
and immunohistochemistry techniques featuring DAB) are available without
charge via the Internet. To date, data are available which represents approx-
imately 70% of the proteins encoded by the human genome (14,079 genes,
17,298 antibodies, and
>
10
7
images).
Skeletal muscle is featured in the Human Protein Atlas. Thus, a goal of
the project is to visualize every protein expressed by human skeletal muscle.
Since human skeletal muscle is most typically a mix of Type I and Type II
fibers, it was of interest to analyze images from the HPA to determine if
fiber-type protein expression could be ascertained, an approach denoted
“histospatial phenotyping.” Interestingly, images of skeletal muscle
from the HPA for proteins with a fiber-specific expression,
including
b
MyHC-
(Type I), MyHC-2A (Type II), and troponin T3 (Type II),
feature either a “striped” pattern, if the muscle was sectioned along the lon-
gitudinal axis of the muscle, or a “checkerboard” pattern if the muscle was
sectioned transversely (
Fig. 7.15
, images in top row). Other patterns which
occur in the image database include images with little or no labeling, which
represent proteins that are likely not expressed at significant levels in skeletal
muscle, or images in which the tissue is labeled to a similar degree, through-
out all of the fibers, indicating that the protein is equally expressed by all fiber
types within the muscle section.
To survey human proteins for fiber-specific expression, a program,
encoded in MATLAB, was used to download the HPA skeletal muscle
image database (
49,000 images, as of this writing). Development of pattern
recognition algorithms that can identify the striped and checkerboard pat-
terns is in progress. Difficulties that have been encountered include a large
variation in the optical density of the images, corresponding to different
degrees of overall labeling by the antibodies and fragmentation of the tissue
microarray cores.
A pragmatic solution was to visually inspect each image. A thumbnail
finder viewer was used for rapidly inspecting and selecting fiber-specific
Search WWH ::
Custom Search