Biology Reference
In-Depth Information
8.
Inspect PCA results. The score plots (samples) and loading
plots (features) are displayed on separate graphs. Default
settings place PC1 on
x
-axis and PC2 on
y
-axis. These can be
changed up to PC5, and up to three PCs can be viewed at once
(in a 3D view that can rotate). The values assigned to each PC
are analogous to the percent of the overall variation that
accounts for each component (see Note 7).
9. If necessary, construct a fi ltered set to perform additional data
reduction (see Note 8). If performed, the analysis is no longer
considered to be unsupervised but can be valuable in visualizing
weak signals by removing background noise (see Subheadings
3.5
and
3.6
).
1.
Import and inspect all gel images.
3.3.2. Quick Guide:
SameSpots
2.
Assign groups and experimental design.
3.
Ensure that the picking references are not included in the anal-
ysis by appropriately using the “Setting Mask of Disinterest”
functionality. This will avoid skewing of the PCA results due to
preferential fl uorescence of the reference markers.
4.
Assess alignment on all images and manually adjust matching
in all images following prescribed methods associated with the
software.
5. Generate normalized ratios on prealigned images (see Note 9).
6. Inspect the PCA results in Progenesis Stats. The score plots
(samples) and loading plots (features) are displayed on a single
overlapping graph (biplot). PCA and hierarchical clustering
(if desired) can be selected under the “Ask another question”
tab. Default settings place PC1 on
x
-axis and PC2 on
y
-axis.
The values assigned to each PC are analogous to the percent of
the overall variation that accounts for each component (see
Note 10).
7.
If necessary, construct a fi ltered dataset by setting tags and
clusters and reevaluate in PCA.
Since PCA organizes independent samples based on the varia-
tion between them,
post hoc
identifi cation of the samples enables
the investigator to establish the relative signal-to-noise ratio (S/N)
present in a dataset. When S/N is high, one can expect proper
organization of the individual samples based on the fi rst two or
three principal components. When S/N is low, additional data
reduction may be necessary to determine if biological variation is
present in the experiment. And when technical noise is extremely
high, the principal components can reveal organization of the sam-
ples that is not based on the biology but rather on some technical
aspect of the sample preparation. What follows are examples of
each of these scenarios, using the PCA functionality of DeCyder
for illustration of the experimental variation.
Search WWH ::
Custom Search