Structural protein interactions predict kinase-inhibitor interactions in upregulated pancreas tumour genes expression data

Micro-arrays can identify co-expressed genes at large scale. The gene expression analysis does however not show functional relationships between co-expressed genes. To address this problem, we link gene expression data to protein interaction data. For the gene products of co-expressed genes, we identify structural domains by sequence alignment and threading. Next, we use the protein structure interaction PSIMAP to find structurally interacting domains. Finally, we generate structural and sequence alignments of the original gene products and the identified structures and check conservation of the relevant interaction interfaces. From this analysis, we derive potentially relevant protein interactions for the gene expression data. We applied this method to co-expressed genes in pancreatic ductal carcinoma. Our method reveals among others a number of functional clusters related to the proteasome, signalling, ubiquitinisation, serine proteases, immunoglobulin and kinases. We investigate the kinase cluster in detail and reveal an interaction between the cell division control protein CDC2 and the cyclin-dependent kinase inhibitor CDKN3, which is also confirmed by literature. Furthermore, our method reveals new interactions between CDKN3 and the cell division protein kinase CDK7 and between CDKN3 and the serine/threonine-protein kinase CDC2L1.