Integrated computational biology analysis to evaluate target genes for chronic myelogenous leukemia

Although hundreds of genes have been linked to chronic myelogenous leukemia (CML), many of the results lack reproducibility. In the present study, data across multiple modalities were integrated to evaluate 579 CML candidate genes, including literature-based CML-gene relation data, Gene Expression Omnibus RNA expression data and pathway-based gene-gene interaction data. The expression data included samples from 76 patients with CML and 73 healthy controls. For each target gene, four metrics were proposed and tested with case/control classification. The effectiveness of the four metrics presented was demonstrated by the high classification accuracy (94.63%; P<2x10^-4). Cross metric analysis suggested nine top candidate genes for CML: Epidermal growth factor receptor, tumor protein p53, catenin ß 1, janus kinase 2, tumor necrosis factor, abelson murine leukemia viral oncogene homolog 1, vascular endothelial growth factor A, B-cell lymphoma 2 and proto-oncogene tyrosine-protein kinase. In addition, 145 CML candidate pathways enriched with 485 out of 579 genes were identified (P<8.2x10^-11; q=0.005). In conclusion, weighted genetic networks generated using computational biology may be complementary to biological experiments for the evaluation of known or novel CML target genes.