Search

Background Aberrant DNA methylation is frequently found in human malignancies including acute myeloid leukemia (AML). While most studies focus on later disease stages, the onset of aberrant DNA methylation events and their dynamics during leukemic progression are largely unknown. Methods We screened genome-wide for aberrant CpG island methylation in three disease stages of a murine AML model that is driven by hypomorphic expression of the hematopoietic transcription factor PU.1. DNA methylation levels of selected genes were correlated with methylation levels of CD34+ cells and lineage negative, CD127-, c-Kit+, Sca-1+ cells; common myeloid progenitors; granulocyte-macrophage progenitors; and megakaryocyte- erythroid progenitors. Results We identified 1184 hypermethylated array probes covering 762 associated genes in the preleukemic stage. During disease progression, the number of hypermethylated genes increased to 5465 in the late leukemic disease stage. Using publicly available data, we found a significant enrichment of PU.1 binding sites in the preleukemic hypermethylated genes, suggesting that shortage of PU.1 makes PU.1 binding sites in the DNA accessible for aberrant methylation. Many known AML associated genes such as RUNX1 and HIC1 were found among the preleukemic hypermethylated genes. Nine novel hypermethylated genes, FZD5, FZD8, PRDM16, ROBO3, CXCL14, BCOR, ITPKA, HES6 and TAL1, the latter four being potential PU.1 targets, were confirmed to be hypermethylated in human normal karyotype AML patients, underscoring the relevance of the mouse model for human AML. Conclusions Our study identified early aberrantly methylated genes as potential contributors to onset and progression of AML. [ABSTRACT FROM AUTHOR]

Background: Classification and variable selection play an important role in knowledge discovery in high-dimensional data. Although Support Vector Machine (SVM) algorithms are among the most powerful classification and prediction methods with a wide range of scientific applications, the SVM does not include automatic feature selection and therefore a number of feature selection procedures have been developed. Regularisation approaches extend SVM to a feature selection method in a flexible way using penalty functions like LASSO, SCAD and Elastic Net. We propose a novel penalty function for SVM classification tasks, Elastic SCAD, a combination of SCAD and ridge penalties which overcomes the limitations of each penalty alone. Since SVM models are extremely sensitive to the choice of tuning parameters, we adopted an interval search algorithm, which in comparison to a fixed grid search finds rapidly and more precisely a global optimal solution. Results: Feature selection methods with combined penalties (Elastic Net and Elastic SCAD SVMs) are more robust to a change of the model complexity than methods using single penalties. Our simulation study showed that Elastic SCAD SVM outperformed LASSO (L1) and SCAD SVMs. Moreover, Elastic SCAD SVM provided sparser classifiers in terms of median number of features selected than Elastic Net SVM and often better predicted than Elastic Net in terms of misclassification error. Finally, we applied the penalization methods described above on four publicly available breast cancer data sets. Elastic SCAD SVM was the only method providing robust classifiers in sparse and non-sparse situations. Conclusions: The proposed Elastic SCAD SVM algorithm provides the advantages of the SCAD penalty and at the same time avoids sparsity limitations for non-sparse data. We were first to demonstrate that the integration of the interval search algorithm and penalized SVM classification techniques provides fast solutions on the optimization of tuning parameters. The penalized SVM classification algorithms as well as fixed grid and interval search for finding appropriate tuning parameters were implemented in our freely available R package 'penalizedSVM'. We conclude that the Elastic SCAD SVM is a flexible and robust tool for classification and feature selection tasks for high-dimensional data such as microarray data sets. [ABSTRACT FROM AUTHOR]