Table 1.

Positive and negative characteristics of different workflows shown in Figure1for the discovery of functional regulatory units

Workflow A	Workflow B	Workflow C
+ ease of application	+ very straightforward, no parameters or thresholds	+ can integrate many existing programs
+ software is available	+ guaranteed result in several rounds	+ different algorithms address particular properties of promoters
+ the spectrum of existing methods covers all particular aspects of transcriptional regulation		+ optimization of a collection of combinatorial modules instead of optimization of each module separately
– big number of methods to choose from (over 150 can be found in the Internet)	– may lead to a scission of a functional module rendering all parts non functional	– huge number of predicted features require much memory and CPU = > specificity filtering should be applied before modules optimization
– relative performance of methods differs for different datasets	– high lab work and time investments
– chance of a correct prediction is ~5-10% [5]
– impossible to estimate the number of required rounds