We are investigating serum microRNAs as biomarkers of islet autoimmunity and type 1 diabetes (T1D). We generated microRNA profiles using LNA RT-PCR arrays from 50 T1D patients, 50 autoantibody-positive relatives (AAb+), and 50 autoantibody-negative relatives (AAb-); 29 trios (T1D, AAb+, AAb-) were from the same family trios and 21 were unrelated. We normalized data and constructed linear mixed models for comparison of expression levels among the groups taking into account the effect of family relationship on variance estimates. False discovery rate (FDR) correction was applied to p values. We identified 33 microRNAs that were differentially expressed among the T1D, AAb+ and AAb- groups: 5 microRNAs, AAb- vs T1D; 23 microRNAs AAb+ vs AAb-; 5 microRNAs AAb+ vs T1D. We used Receiving Operating Curves (ROC) to calculate area under the curve (AUC) values for these microRNAs, which ranged between 0.64-0.70. We then applied predictive models that incorporate multiple microRNAs. Expression levels were classified into categories (high or low) based on observed data. Predictive models of group, considering AAb+ vs AAb- first, and then T1D vs AAb-, were constructed using a stepwise logistic model fitting procedure. The significance of the associations were determined based on these predictive models, as well as the overall predictive ability of the model based on ROC curves constructed from model predicted probabilities and the associated AUC. Two combinations of microRNAs were significantly associated with group: let-7g, miR-222, miR-1183, and miR-141*, and miR-9, miR-222, miR-1183, and miR-141*. Each of these two combinations yielded improved AUC values (combined AUC range 0.74-0.80 vs 0.64-0.70 range for individual microRNAs). Of note, several of the microRNAs identified have reported functional links with beta cell function (miR-9) and various immune pathways and autoimmune conditions. Analysis of larger cohort (ongoing) will further inform on which combinations of serum microRNAs can aid T1D prediction.