March 24, Wed 2010
11:00 am, MRB 200 Conference Room
Dr. Mihaela Sardiu
Stowers Institute for Medical Research
Towards a comprehensive understanding of protein networks underlying chromatin remodeling complexes
Protein complexes are major building blocks of biological systems. Therefore, when aiming to understand biological events on a systems level, it is essential to identify and characterize protein complexes, and the interrelationships among proteins through parameter quantification, followed by bioinformatics analysis and mathematical modeling.
Towards this goal, we tested the feasibility of using quantitative proteomics data generated from APMS (affinity purification followed by mass spectrometry) experiments in characterizing dynamic protein complexes. We applied a combination of clustering techniques, machine learning, vector algebra and information theory in order to characterize protein complexes and to determine the inter-dependency of its subunits/ proteins. Using data sets generated from protein complexes involved in chromatin remodeling and transcription, we could demonstrate and experimentally verify that quantitative proteomics can be used for generating probabilistic protein networks. However, since many protein complexes are very stable, the value of this approach is limited. Therefore we developed approaches to overcome this problem using either genetic perturbations or, in the case of protein complexes that assemble and function in the nucleus, by using the cytoplasmic fraction in order to determine the interrelationships/ interdependencies of subunits of protein complexes.
A striking property of the quantitative deletion interaction networks generated from two protein complexes, yeast Rpd3 and SAGA, is that these models provide novel information about the function of proteins, based upon a similar topology within the complexes.