Mathematical and computational analysis of biological networks uncovers evolution, disease, and gene functions

by Prof. Nataša Pržulj (Department of Computing, Imperial College London and Računarski Fakultet, Belgrade)

Wednesday 29 Dec 2010, 17:00 → 18:00 Europe/Belgrade

1 (Računarski Fakultet)

Genes produce thousands of different protein types that interact in complex networked ways and make cells work. It is possible that interaction network data will be as useful as the sequence data in uncovering new biology. Given the abundance of interaction data, systems-level comparisons of networks of pathogenic and non-pathogenic species could play a vital role in understanding mechanisms of pathogenicity. Also, comparing networks of healthy and disease-affected cells could deepen our understanding of disease and lead to identification of cellular parts that are candidates for therapeutic intervention. Furthermore, biological network comparison and alignment could enable transfer of knowledge between species, since we may know a lot about bio-molecules in one species and almost nothing about aligned bio-molecules in another species.

Existing network alignment methods use information external to network topology, e.g., sequence data.  Since network topology provides a new and independent source of biological information, it is important to understand how much biology we can learn from it independently from any other data source.  Hence, we develop mathematically rigorous ways for aligning networks based solely on their topology. Our network aligners produce by far the most complete alignments of biological networks to date, exposing large and contiguous regions of network similarity even for as distant species as yeast and human thus suggesting broad similarities in internal cellular wiring across all life on Earth.  Moreover, they demonstrate that protein function and species phylogeny can be extracted solely from network topology. Analogous to reconstruction of phylogenetic trees using sequence similarities, we use our network alignment similarities to successfully reconstruct the phylogenies of protists, fungi, and herpesviruses.

In addition, we show that network topology around cancer and non-cancer genes is different and use this to predict new cancer genes. Our predictions are phenotypically validated. We present evidence that topology-based analyses of biological networks provide new biological and phylogenetic insights and that they can help identify novel drug targets, hence aiding therapeutics and health care.