Differential Networks: Dynamics, Alignment, and Evolution

Although cellular behaviors are dynamic, the networks that govern these behaviors have been mapped primarily as static snapshots. It is still largely unknown to what extent network architecture is perturbed by genetic, environmental, or evolutionary changes. To address this question, we have been working on a number of approaches to align and compare large biological interaction maps.

The first set of approaches is for comparing interaction networks across multiple species [Bandyopadhyay et al. Genome Res. 16(3):428-35 2006; Sharan et al. ProcNatlAcadSci8:102(6) 1974-79 2005; Kelley et al. ProcNatlAcadSci100, 11394-9 2003]. Some of these methods have been made available as the network alignment tools PathBLAST and NetworkBLAST [Kelley et al. Nucleic Acids Research 1;32: W83-8 2004; Kalaev et al. Bioinformatics 24(4):594-6 2008]. These tools have application to infectious disease, for example by targeting drugs to pathways that are present in a pathogenic organism but absent from its human host [Suthram et al. Nature438(7064):108-12 2005].

A second focus has been to develop methods to study how networks are remodelled by environmental conditions. For this purpose we have developed an approach we call differential epistasis mapping, or dE-MAP [Bandyopadhyay et al. Science330(6009):1385-1389 2010]. The core concept is that networks should not be analyzed individually in a single condition, but that they should be generated for at least two different conditions and quantitatively subtracted one

Differential epistasis mapping

Differential epistasis mapping [Bandyopadhyay et al. Science 2010]

Network alignment of Plasmodium vs. Yeast

Network alignment of Plasmodium (left) vs. yeast (right) [Suthram et al. Nature2005]

from the other. This process reveals a new “differential” network that contains many interactions that go undetected in either static condition alone. Moreover, differential epistasis mapping identifies the precise set of interactions that function in the response to perturbation, subtracting away all other (less relevant) interactions.

A current roadblock to advancing Comparative Network Biology as a field is the marked lack of interaction maps at high coverage and at the appropriate distances for evolutionary comparison. To address this shortcoming, we are working with the Krogan lab at UCSF to obtain high-density physical and genetic interaction maps across the model organisms Schizosaccharomycespombe and Saccharomyces cerevisiae [Roguev et al. Science. 322:405-410 2008]. The last common ancestor of S. pombe and S. cerevisiae is quite ancient (at least 400 mya), making the conserved interaction map generalizable to large parts of the eukaryotic lineage. Many aspects of S. pombe physiology bear more in common with mammals than does S. cerevisiae, including intron/exon splicing, chromosomal architecture, and RNA interference machinery. To define an appropriate scope for the interaction mapping efforts, our current screens are targeted to a set of ~400 proteins covering the majority of kinases and DNA-binding transcription factors. Comparative network analysis in the lab is funded by grant GM084279 from the National Institute for General Medical Sciences and grant IIS-0803937 from the National Science Foundation.