d using the limma package with a critical fold change value 0.5
ical p value 0.01. This led to 910 DMSs to be used. The reason
hese specific critical values was for generating roughly balanced
d DMSs for the affordable analysis in a normal computer.
wards, two objectives were investigated for this data set. The first
nvestigate how the differential expressions of each DEG were
ed or determined by the differential methylations of 910 DMSs.
called the methylation-to-expression pattern discovery. This set
ls were called the M2E models. A regression model was
ed between one DEG and all 910 DMSs which were used as the
s. After the quantitative relationship between DEGs and DMSs
en modelled using the regression analysis models, the most
t DMS contributor of each DEG was identified and analysed.
r, whether the differential methylation of the local methylation
the remote methylation sites was the top contributor to the
al expression pattern of a gene was then examined.
econd aimed to discover the expression-to-methylation pattern.
of models were named as E2M models. For each DMS, a
n model was constructed between its differential methylations
differential expressions of 1,250 DEGs. Through this regression
e quantitative relationship between the differential methylations
ifferential expressions was used to identify the most influencing
a DMS. Similarly, whether the differential expression of the local
the remote gene was the major contributor to the differential
on of a methylation site was examined.
main aim of the study was not the simple ranking of the
ent variables or regressors in both M2E and E2M models. Instead,
anked regressor of each regression model, which was either a
a DEG, was discovered and analysed in terms of their genomic
hip with their target variable, which was either a DEG or a DMS.
ess of these models was to examine whether a local regressor or
regressor is the main contributor or the determinant of a target
Therefore, this process was also referred to as an impact analysis.