Systems Biology (Methods in Molecular Biology)

(Tina Sui) #1
The same “balance” character is present in PC3. The two genes
most affected by PC3 (higher loadings in absolute value) are PolB
and Mlh1. These two genes have opposite sign correlations with
PC3, this implies that this “hidden” pathway encompasses a bal-
ance between these two genes expression: high values of PolB
correspond to low values of Mlh1 and vice versa that corresponds
to our initial hypothesis. From what we said before, this negative
correlation between the two genes is not true in every context and
thus does not appear in terms of direct correlation coefficient
between the relative expression level as such, this balance only
holds as for the specific pathway correspondent to PC3.
Genes are “probes” of processes (components) working “on
behind,” thus the actual values of expression can be expressed as a
function of component scores. In Fig.3 the PolB expression values
are modeled as a weighted summation of component scores (vector
points are patients,Yvalues the experimentally observed expression
values of PolB, while the abscissa corresponds to the PolB values
estimated as weighted summation on components).

Fig. 3The observed values of PolB expression level come from a weighted summation of the contributes of the
first three principal components. The PolB expression is a mixture of 33 parts of PC1, 9 parts of PC3, and
42 parts of PC3. Each component corresponds to an independent “regulation flux” of DNA repair response,
genes are not the “actual players” but probes of underlying hidden factors


66 Alessandro Giuliani

Free download pdf