Systematically validated inference of quantitative regulatory networks and condition-specific TF activity
•Correct direction of perturbation:
In samples where TFs are deleted/induced, percent correctly inferred activity values as less/greater than in the WT sample
•Median rank percentile of perturbed TF:
In samples where TFs are deleted/induced, median rank of inferred activity levels compared to unperturbed TFs. Rank 1 = 100%
•Positive TFA-mRNA correlation:
Percent of TFs whose inferred activity positively correlates with mRNA levels of their encoding genes
Additional metrics used to emphasize the benefits of optimizing a quantitative regulatory network together with TF activity:
•Percent of literature supported edges identified:
In samples where a known regulator of TFA are deleted, the percent of TFs with a change in activity above a rank threshold
•Percent of fits in the correct direction:
In time-courses where TF-encoding genes are induced, the percent of TFs whose pattern of activity can be fit to an increasing sigmoid model above a variance explained threshold