Gene expression and risk of leukemic transformation in myelodysplasia.
Yusuke ShiozawaLuca MalcovatiAnna GallìAndrea PellagattiMohsen KarimiAiko Sato-OtsuboYusuke SatoHiromichi SuzukiTetsuichi YoshizatoKenichi YoshidaYuichi ShiraishiKenichi ChibaHideki MakishimaJacqueline BoultwoodEva LindbergSatoru MiyanoMario CazzolaSeishi OgawaPublished in: Blood (2017)
Myelodysplastic syndromes (MDSs) are a heterogeneous group of clonal hematopoietic disorders with a highly variable prognosis. To identify a gene expression-based classification of myelodysplasia with biological and clinical relevance, we performed a comprehensive transcriptomic analysis of myeloid neoplasms with dysplasia using transcriptome sequencing. Unsupervised clustering of gene expression data of bone marrow CD34+ cells from 100 patients identified 2 subgroups. The first subtype was characterized by increased expression of genes related to erythroid/megakaryocytic (EMK) lineages, whereas the second subtype showed upregulation of genes related to immature progenitor (IMP) cells. Compared with the first so-called EMK subtype, the IMP subtype showed upregulation of many signaling pathways and downregulation of several pathways related to metabolism and DNA repair. The IMP subgroup was associated with a significantly shorter survival in both univariate (hazard ratio [HR], 5.0; 95% confidence interval [CI], 1.8-14; P = .002) and multivariate analysis (HR, 4.9; 95% CI, 1.3-19; P = .02). Leukemic transformation was limited to the IMP subgroup. The prognostic significance of our classification was validated in an independent cohort of 183 patients. We also constructed a model to predict the subgroups using gene expression profiles of unfractionated bone marrow mononuclear cells (BMMNCs). The model successfully predicted clinical outcomes in a test set of 114 patients with BMMNC samples. The addition of our classification to the clinical model improved prediction of patient outcomes. These results indicated biological and clinical relevance of our gene expression-based classification, which will improve risk prediction and treatment stratification of MDS.
Keyphrases
- gene expression
- bone marrow
- machine learning
- dna methylation
- dna repair
- deep learning
- induced apoptosis
- end stage renal disease
- signaling pathway
- genome wide
- poor prognosis
- ejection fraction
- single cell
- newly diagnosed
- mesenchymal stem cells
- acute myeloid leukemia
- prognostic factors
- oxidative stress
- cell death
- big data
- epithelial mesenchymal transition
- endoplasmic reticulum stress
- copy number
- smoking cessation
- binding protein
- patient reported outcomes
- data analysis
- artificial intelligence
- transcription factor
- bioinformatics analysis
- replacement therapy