Bayesian Network Inference Algorithms - Bayesian Networks in R: With Applications in Systems Biology - page 98

> dbn2.fit[["265768_at"]] = lasso.t
> dbn2.fit[["245094_at"]] = lasso.s
Using the fitted network dbn2.fit , we can now call both cpquery and cpdist
to perform smoothing, filtering, and prediction. We may be interested, for example,
in the inhibitory effects of prolonged high levels of expression of gene 245094 at
(at times t − 2and t − 1) on gene 265768 at .

> cpquery(dbn2.fit, event = ('265768_at' > 8),
+ evidence = ('245094_at' > 8) & ('245094_at1' > 8))
[1] 0.1554545
This probability is much lower than the corresponding probability computed con-
ditioning only over time t

1( 0.2448749 ), supporting the hypothesis that the
inhibitory effects of 245094 at are protracted over time.
This hypothesis is also supported by the fact that, regardless of the expression lev-
els of gene 245094 at at time t

−

− 1, conditioning on high expression levels at time

t 2 results in a much lower probability of high expression of gene 265768 at at

time t (compared to the unconditional probability, computed by setting evidence
= TRUE ).

−

> cpquery(dbn2.fit, event = ('265768_at' > 8),
+ evidence = ('245094_at1' > 7) & ('245094_at1' < 8))
[1] 0.9555499
> cpquery(dbn2.fit, event = ('265768_at' > 8),
+ evidence = TRUE)
[1] 0.4507846
Finally, given our knowledge on gene 265768 at and gene 245094 at at
times t − 2and t , we can investigate the distribution of 245094 at at time t − 1.

258736_at 257710_at 255070_at 245319_at 245094_at1

245094_at

265768_at

Fig. 4.4 Dynamic Bayesian network for the expression levels of gene 265768 at going back
from time t to time t − 2