Inferring gene regulatory networks - evaluation

Question

[Zara]

Hi everyone,

I'm using DREAM 4 challenge data in order to evaluate my gene regulatory inference method.

I have a couple of questions:

1. How should I use the perturbation data. I don't know what does G1(-/-) or G1(+/-) mean. Does this plus and minus signs represent mutations? How come we have two sign separated with a slash?
2. In several papers^* I read that a linear regression model (or something close to it) has been quite sufficient for network inference. I don't get any good result with a sparse linear regression model. Am I missing something?

^* for example the model proposed by Inferelator is almost linear, the models described in http://mlg.eng.cam.ac.uk/pub/pdf/KnoGha11b.pdf use a weighted sum as the effect of genes on each other.

However I could not get good results using this weighted sum function. I tested several datasets (DREAM 4 and 3 in silico data, SOS network data, ... )

Can anyone help me in this case?

Any help is appreciated.