3.2. Effect of active threshold
Selecting active threshold significantly influence the prediction
precision of essential proteins. Apparently, the reliability of the
dynamic PPI network evidently increases with increasing active
threshold. As shown in Table 1, the average GO similarity of the
dynamic PPI network increases with increasing active threshold.
However, higher active threshold means fewer proteins and fewer
essential proteins contained in the dynamic PPI network (Fig.1). The
amount of proteins and essential proteins in the dynamic PPI
network slowly decreases when the active threshold increases from
0 to 0.8 but rapidly decreases when the active threshold increases
from 0.8 to 0.99. This result indicates that 0.8 is a demarcation point.
Moreover, the dynamic PPI network still contains more than 93% of
essential proteins in the static PPI network when the active threshold
is 0.8. Therefore, the active threshold is set to 0.8 for the construction
of a relatively reliable dynamic PPI network with enough proteins.
This relatively high threshold is in accordance with what has been
pointed out by Wang et al. (2013) that for each protein, the time
points with its highest expression value and with expression value
near to the highest one should be regarded as its active time points.