Accuracy of Predicting IgHV Mutation Status in Chronic Lymphocytic Leukemia Using RNA Expression Profiling and Machine Learning

Background: Immunoglobulin heavy chain (IgHV) mutation status is a unique prognostic indicator for predicting clinical course and response to targeted therapy in chronic lymphocytic leukemia (CLL). Most of the tests for the IgHV mutation status require extensive and complicated genomic evaluation. We explored the potential of using RNA expression data generated from routine targeted RNA sequencing by Next Generation Sequencing (NGS) along with machine learning for the prediction of the IgHV mutation status. Methods: NGS is used first to sequence IgHV DNA and to determine mutation status of 120 CLL samples. The RNA of these samples was sequenced using targeted panel of 1,408 genes. Geometric Mean Naïve Bayesian (GMNB) was used to select genes that distinguish between mutated and unmutated. Machine learning algorithm then used to predict the IgHV mutation status. Results: The algorithm showed a receiver operating characteristic curve with area under the curve (AUC) of 0.927. A sensitivity of 86% (95% confidence interval: 74.5–93%) and specificity of 93% (95% confidence interval: 82–98%) were achieved in distinguishing between the IgHV mutated and unmutated. Validation using leave one out showed AUC of 0.870. Blind testing of 22 additional CLL samples showed 91% concordance between IgHV mutation status as detected by DNA sequencing and mutation status as predicted by RNA and machine learning algorithm. The selected top 23 genes used in this machine learning model included growth factors, transcription factors, and oncogenes. Conclusions: This data demonstrates that RNA expression when combined with a machine learning algorithm can reliably predict IgHV mutation status with high sensitivity and specificity. This approach is simple and not dependent on the purity of the isolated CLL clone. Furthermore, this approach defines specific genes that are crucial in distinguishing between mutated and unmutated CLL.