Under the guidance of Roded Sharan, along with Omer Tamuz, I worked on several problems in systems biology.
We created the PROtein Module Optimizer (PROMO) which is a method to identify modules in a Protein-Protein Interaction (PPI) network. Given a PPI network we first measure the likelihood of all pairs of proteins to interact using the Log-Likelihood statistic. We then receive a full graph G' with positive and negative weights on its edges. Finding a module in the PPI network thus translates to finding a subset of vertices U in G' in which the sum of the weights on the edges between the vertices in U is maximal. This problem is computationally hard. We use a branch-and-bound strategy to solve this problem on the network. While a naive branch-and-bound scheme fails to provide a solution on the PPI networks, PROMO has various improvements which enable it to obtain the optimal solution. The main improvement is its ordering of the vertices in the branching process according to their (estimated) expectancy to improve the solution found thus far.
For more details you can check out the PROMO paper .
We also worked on identifying modules through comparative analysis. That is, given a module in one PPI network (e.g. yeast), finding a homologous module in another PPI network (e.g. fly). In this method we also allow applying several biological constraints on the modules identified.
In the meantime you can check out the presentation on comparative analysis.