NetworKIN Depends on What You Do, Where You Are, and Who You Know

Science's STKE  03 Jul 2007:
Vol. 2007, Issue 393, pp. tw231
DOI: 10.1126/stke.3932007tw231

Signaling networks regulate cell behavior by means of posttranslational modifications that alter protein function. Phosphorylation, perhaps the most familiar of these modifications, constitutes a widespread regulatory mechanism: There are more than 500 human kinases, and thousands of phosphorylation sites have been identified. The limited specificity of phosphorylation consensus motifs, however, together with contextual factors that influence the likelihood that two proteins interact, have constrained our ability to determine precisely which protein phosphorylation sites constitute substrates for particular kinases. Linding et al. developed a computational approach, which they called NetworKIN, that combines predictions of kinase families likely to phosphorylate particular motifs with contextual information on substrates and kinases to predict which kinases mediate phosphorylation of specific sites. Accuracy in predicting the correct kinase family involved in phosphorylating a test group of known sites using contextual information was substantially improved compared with predictions based on consensus motifs alone. Indeed, in some cases, contextual information enabled prediction of the specific kinase involved. Application of the NetworKIN algorithm led to novel predictions that were confirmed experimentally. For instance, Rad50, part of a complex implicated in stabilizing broken chromosomes, was identified as a substrate of the kinase ATM (ataxia telangiectasiamutated), which is activated in the DNA damage response, and 53BP1 was identified as a substrate for CDK1 (cyclin-dependent kinase 1). Combining the algorithm with quantitative mass spectrometry—a methodology potentially applicable to large-scale mapping of phosphorylation networks—suggested that the transcriptional repressor BCLAF1 (BCL2-associated transcription factor 1) is a substrate for GSK-3 (glycogen synthase kinase 3). Thus, including contextual information should facilitate modeling in vivo phosphorylation networks.

R. Linding, L. J. Jensen, G. J. Ostheimer, M. A. T. M. van Vugt, C. Jørgensen, I. M. Miron, F. Diella, K. Colwill, L. Taylor, K. Elder, P. Metalnikov, V. Nguyen, A. Pasculescu, J. Jin, J. G. Park, L. D. Samson, J. R. Woodgett, R. B. Russell, P. Bork, M. B. Yaffe, T. Pawson, Systematic discovery of in vivo phosphorylation networks. Cell 129, 1415-1426 (2007). [PubMed]