Uncovering Differential Multi-microRNA Signatures of Acute Myeloid and Lymphoblastic Leukemias with a Machine-Learning-Based Network Approach

摘要: Complex phenotypic differences such as those among different acute leukemias cannot be fully captured by analyzing the expression levels of one single microRNA at a time. We introduce data-driven approach that identifies which and how many microRNAs are needed to differentially characterize myeloid (AML) lymphoblastic (B-ALL T-ALL) leukemias. First, global human was measured on AML, B-ALL, T-ALL cell lines patient samples. Then, systematically applying support vector machines all taken pairwise, we built an AML-centric microRNA-dyad network based only line data. These dyads were able classify samples very reliably (accuracy >94%). In order B-ALL T-ALL, however, least 3 needed. thus microRNA-triad ensembles using scalable framework preselects top-scoring GenePattern class prediction method. validated our findings some well-characterized microRNAs. For instance, members miR-23a cluster (which includes also miR-24 miR-27a), known function tumor suppressors leukemias, appeared in centric networks playing similar roles from network-topological perspective. The most connected B-ALL-centric miR-708 (70% triads). Microarray qRT-PCR analyses across showed is highly specifically expressed suggesting might serve biomarker for B-ALL. However, experimentally manipulating lines, no functional effect revealed survival or proliferation, underscores cooperative nature cellular involving multiple