Atypical chronic myeloid leukemia (aCML) and chronic myelomonocytic leukemia (CMML) represent two histologically and clinically overlapping myelodysplastic/myeloproliferative neoplasms. Also the mutational landscapes of both entities show congruencies. We analyzed and compared an aCML cohort (n = 26) and a CMML cohort (n = 59) by next-generation sequencing of 25 genes and by an nCounter approach for differential expression in 107 genes. Significant differences were found with regard to the mutation frequency of TET2, SETBP1, and CSF3R. Blast content of the bone marrow revealed an inverse correlation with the mutation status of SETBP1 in aCML and TET2 in CMML, respectively. By linear discriminant analysis, a mutation-based machine learning algorithm was generated which placed 19/26 aCML cases (73%) and 54/59 (92%) CMML cases into the correct category. After multiple correction, differential mRNA expression could be detected between both cohorts in a subset of genes (FLT3, CSF3R, and SETBP1 showed the strongest correlation). However, due to high variances in the mRNA expression, the potential utility for the clinic is limited. We conclude that a medium-sized NGS panel provides a valuable assistance for the correct classification of aCML and CMML.
Keywords: CMML; NGS; aCML; machine learning algorithm; nCounter.
© 2019 The Authors. Cancer Medicine published by John Wiley & Sons Ltd.