Impact of Noise on Molecular Network Inference
Figure 3
Bayesian networks inferred using Grow-Shrink algorithm along with Pearson correlation ( = 0.01) for the three-gene network motifs, namely: common-effect (a-c), three-chain (d-f) and coherent type-I feed-forward loop (g-i) for various choices of noise variances:
and
.
The confidences of the edges are represented as percentage of the edges that persisted across 200 independent realizations. Edges with
are shown by solid arrows whereas others
are shown by dotted arrows. Edges with confidence
are deemed noisy and excluded for clarity.