通过将活性与药物靶点邻近性关联来验证上游激酶预测

Introduction: Sustained cell signaling is one of the hallmarks of tumor growth. Deregulation of kinase signaling can be studied by methods such as peptide microarrays or phospho-proteomics. For this, the prediction of kinases from phosphorylation signatures is a critical and complex. The aim of our study is to validate the biological relevance of kinase predictions, which we evaluate by the integration of kinase activity profiles with sensitivity data from the Genomics of Drug Sensitivity in Cancer (GDSC). A fundamental sensitivity mechanism involves repressing the activity of the drug's target kinase and its downstream survival pathways. We can test this mechanism through network analysis, hypothesizing that the signaling networks of active kinases and drug targets show high connectivity in sensitive cells, that should reflect the biological relevance of the kinase predictions. Methods : Serine/Threonine Kinase (STK) activity profiling of 11 B-cell lymphoma cell lines was performed via the KinomePro platform (PamGene International B.V.). Kinases were predicted for 10 cell lines generally sensitive to multiple drugs (IC 50 < 1 µm) compared to one relatively resistant control line (showing sensitivity to a smaller number of drugs). Cell line specific networks were generated from the top kinases and the target kinases of drugs the cells were most sensitive to, using the STRING protein-protein interaction database and Prize-Collecting Steiner Forest algorithm. Results: To quantify signaling proximity of kinases and drug targets, we developed the network connectivity score. For each cell line, we calculated network connectivity using the median of the shortest paths between each kinase and drug target. This was then tested for statistical significance against a reference set of 50 random networks, generated using the original drug targets and randomized kinase data. Kinase predictions for 5 out of 10 B-cell lymphoma cell lines resulted in significant network connectivity score (p < 0.07), depending on the parameters used in the prediction algorithm. These results demonstrate the method's utility to identify optimal parameter settings for the prediction algorithm. Conclusion: We developed a validation method for the biological relevance of kinase predictions from phosphorylation signatures obtained in a cellular context. Future work will focus on applying this method to improve studies of the role of kinases in signal transduction by evaluating and optimizing the performance of kinase prediction methods and their potential biases.