Application of principle component analysis and logistic regression in analyzing miRNA markers of brain arteriovenous malformation

Abstract

Brain arteriovenous malformation (BAVM) is frequently described as vascular malformation. Although computer tomography (CT), magnetic resonance imaging (MRI) and angiography can clearly detect lesions, there are no diagnostic biological markers of BAVM available. Current study demonstrated that microRNA (miRNA) showed a feasible marker for vascular disease. To find key correlations between these miRNAs and the onset of BAVM, we carried out chip analysis of serum miRNAs by identifying 18 potential markers of BAVM. We then constructed a principle component analysis and logistic regression (PCA-LR) model to analyze the 18 miRNAs collected from 77 patients. Another 9 independent samples were used to test the resulting model. The results showed that miRNAs hsa-mir-126-3p and hsa-mir-140 are important protective factors, while hsa-mir-338 is a dominating risk factor, all of which have stronger correlation with BAVM than others. We also compared the testing results using PCA-LR model with those using LR model. The comparison revealed that PCA-LR model is better in predicting the disease.