Association of FLT3 and NPM1 mutations in patients with acute myeloid leukemia with metabolomic patterns determined by mass spectrometry.

Abstract

Listen

Translate article

e19003 Background: Patients diagnosed with Acute Myeloid Leukemia (AML) are categorized into risk groups based on cytogenetic and molecular abnormality tests, which determine the specific treatment regimes. Since mutation tests take time and some emergency patients require immediate treatment, a method to provide fast clinical data that will be the basis for initial treatment regime is needed. Aim of the study is to discover patterns associated with FLT3 and NPM1 mutations in AML patients, which will provide a fast clinical information to clinicians. The hypothesis is based on the previous studies which indicate a link between NPM1 and FLT3 mutations with metabolism. The mutations were chosen based on their high frequency in AML patients and their critical role in risk group determination. Methods: All AML patients (n = 42) and healthy individual samples (n = 16) were obtained from Ankara University Hematology Department. Patient groups were divided into five groups based on the mutation status. Pattern determination of the mutations was achieved by LC-MS/MS measurements of amino acid and acyl carnitine panels that are highly associated with metabolism. After preprocessing of the raw data, univariate (ANOVA) and multivariate analyses (PCA and PLS-DA) were performed to define class discrimination between sample groups. The remarkable analytes that have significant power in the discrimination were determined by VIP analysis. The developed model was validated with K-Fold cross validation method and permutation test. The most significant pathways in class discrimination were identified with pathway analysis. Visualization was accomplished via Metaboanalyst 5.0 software. Results: Principal Component Analysis (PCA) showed that 79% of the total variance of the sample groups was explained by the model. In order to increase class discrimination, Partial Least Squares-Discriminant Analysis (PLS-DA) was performed. R2Y and Q2 were found as 0.845 and 0.619, respectively. PLS-DA model was validated with K-fold analysis and permutation test. In all the validation experiments, a low cross-validation error was observed. In VIP analysis, the most significant analytes that cause the class discrimination were found as C0 carnitine, glutamic acid, aspartic acid, respectively. In the pathway enrichment analysis performed with these analytes, aminoacyl t-RNA biosynthesis, arginine biosynthesis, valine-leucine-isoleucine biosynthesis were found as statistically significant pathways responsible for the class discrimination. Conclusions: A preliminary model based on the targeted metabolomics approach was developed for the prediction of mutation status of NPM1 and FLT3 proteins in AML patients. Proposed model has a high fit value, validity, and strong predictive power. The reliability and validity of the model can be further increased by future multicenter studies.