The presence of factors external to genes makes genetic expression variable. Guziolowski and others propose a way to model a biomolecular network through the use of logical rules imposed onto a network of some known interactions. They propose two algorithms to analyze the triggering of inhibitions or activations of some molecules, given the rest of the molecular environment. The first recovers a minimal subgraph that connects some observed values and finds possible pathways for activation or inhibition of a given target. This algorithm does not evolve in polynomial time. The second algorithm finds some unobserved or unpredicted nodes by artificially fixing one, and then finding contradictions in the resulting subgraphs; this makes it possible to find the subgraph of the pathway leading to the proper unpredicted node.

The advantage of using logical rules imposed onto a network, as opposed to statistical microarrays, is that the latter suffer from the generation of many false positives when thousands of genes are tested. In their algorithms, the authors modify the prediction probabilities, biasing them through logical (molecular) rules for activation or inhibition.

The algorithms in this paper have been validated for pathways that include the correlation of an oncogene with its inhibition or activation in the presence of some other molecules. The results are quite encouraging.

I recommend this paper to molecular biologists, soft computing and artificial mathematicians, and related professionals. Some advanced graduate students might be able to grasp the content as well.