Due to recent sensory technology advancement, there is a huge surge in metabolomics, proteomics, glycomics, and genomic datasets. This opens the door to large comparative studies and inference that can shed light on the underlying dynamics and facilitators of the many organic processes.

We develop statistical methodologies for studying such data and robustly interpreting the    resulting models. The statistical methodology developed is based on advanced computational algorithms, signal processing, machine learning and robust statistics methods.

In particular, we concentrate on time series proteomics and study data size effect on graph model estimation such as causal models and Bayesian models.


Graphical result of amino acid mutations associated with different sub-types of HIV conditioned by Nelfinavir (NFV) treatment. Red edges suggest exceptionally significant associations with high IRR and  scoring. Edge directions are such that the presence of the edge source will increase the probability of the edge target.
Taken from Bar-Yaakov et al., J. of Computational Biology, 2012