A microRNA Expression-Based Model Predicts Event Free Survival in Pediatric Acute Myeloid Leukemia

INTRODUCTION Acute myeloid leukemia (AML) is a hematopoietic malignancy that comprises almost 25% of pediatricleukemiasand is characterized by genetic and epigenetic alterations that lead to impairment of differentiation and clonal expansion. Despite intensive chemotherapy, more than half of children with AML either fail to achieve a complete remission (CR) or relapse after initial response.As such, the availability of a predictor of treatment failure at diagnosis may allow early institution of alternative therapies to improve outcome. MicroRNAs (miRNAs), a class of small non-coding RNAs that regulate the translation of their target mRNAs, are frequentlydysregulatedin cancers, and thus may serve as robust biomarkers of patient outcome. METHODS To identifymiRNAbiomarkers associated with treatment failure and candidate therapeutic targets, we performed a comprehensive sequence-based characterization of the pediatric AMLmiRNAexpression landscape usingmiRNA-seqdata from 637 primary samples. AmiRNAexpression-based model for EFS separating patients into low, intermediate, andhigh riskgroups was produced using penalized Cox regression. The model was designed usingmiRNAexpression data obtained from a training cohort, which consisted of two-thirds of our study cohort (n=425), and then tested on the remaining one-third of our study cohort (n=212). The training and test cohorts were derived by random selection. RESULTS A 36-miRNA EFS predictive model was generated. This model was comprised of16miRNAsthat were over-expressed and 20 that were under-expressed in patients who experienced an event (death, relapse or IF). Among the 36miRNAtranscripts were miR-155, miR-335, miR-139 and miR-375, which have been previously individually associated with survival in pediatric AML. To demonstrate the potential clinical utility of the model, we determined 2 miRNAmodel score thresholds in the training cohort to separate patients into low, intermediate and high miRNAmodel score risk groups. The miRNAmodel score groupings were significantly associated with EFS in both the training cohort and test cohort (Figure 1A; P P P Furthermore, to demonstrate the strength of our predictive model, we evaluated the clinical significance of the model in each of the low, standard and high CM risk cohorts. In this analysis, our model was capable of further stratifying patients in each of the 3 CM risk cohorts into 3 distinct miRNAmodel score-based risk categories (Figure 1B, P CONCLUSIONS We present amiRNAexpression-based predictor of outcome in pediatric AML, comprised of 36miRNAtranscripts. Our predictive model was developed and tested on a large cohort of primary patient samples (n=637), and demonstrated that diagnosticmiRNAexpression profiles can identify risk groups in patients independent of other CM risk factors. Moreover, this model is applicable to RNA from samples that are routinely obtained for diagnosis, and thus has the potential to impact clinical practice. Disclosures No relevant conflicts of interest to declare.