The MDS Hub uses cookies on this website. They help us give you the best online experience. By continuing to use our website without changing your cookie settings, you agree to our use of cookies in accordance with our updated Cookie Policy

Targeted sequencing of seven genes may improve classification of MDS

Feb 10, 2021

The National Myelodysplastic Syndromes (MDS) Natural History Study (NCT02775383) is a collaborative study involving the National Heart, Lung, and Blood Institute (NHLBI), National Cancer Institute (NCI), and NCI Community Oncology Research Program (NCORP). It is currently investigating whether targeted exon sequencing may be useful to increase compliance between local and central pathologic review of diagnosis, by analyzing the clinical, genetic, and epigenetic characteristics related to the onset and evolution of the disease in patients with newly diagnosed or suspected MDS.

The results were presented by Johannes Goll during the 62nd American Society of Hematology Annual Meeting and Exposition,1 and here, we are pleased to summarize the key points of his talk.


  • The National MDS Natural History Study is a prospective, observational cohort study:
    • Conducted at > 120 centers
    • With an enrolment target of up to 2,000 patients with MDS or MDS/myeloproliferative neoplasms (MPN) overlap, and up to 500 patients with idiopathic cytopenias of unknown significance (ICUS)
  • Peripheral blood and bone marrow samples collected from untreated, cytopenic patients and previously examined by local pathologists were reviewed by central pathologists; an independent review was used if there was a conflict
  • Diagnosis classifications were as follows:
    • MDS*
    • MDS/MPN overlap
    • Acute myeloid leukemia (AML) blasts < 30%
    • Other: clonal cytopenia, morphologic dysplasia, AML blasts ≥ 30%, other myeloid malignancies
  • The agreement between local and centralized pathologic review was examined against these diagnostic classifications
  • The largest absolute discrepancies were observed between the diagnosis of MDS versus Other
  • Therefore, the goal of this study was to assess whether targeted gene sequencing could be used to increase agreement between pathologic diagnoses in these two categories.

*The updated 2016 World Health Organization (WHO) classifications were followed to diagnose MDS in the central review.


A total of 648 patients (MDS, n = 212; Other, n = 436) were included in the analysis. There were 90 patients with clonal cytopenia of undetermined significance (CCUS) and 89 patients with other cancers in the Other category.

Targeted exon sequencing was performed in 96 myeloid genes with a mean coverage of 1,317X. The threshold for minimum variant allele frequency (VAF) was 2%. To create the model and perform the first test, likely disease-causing variants were manually reviewed in 596 subjects (Figure 1).

Figure 1. Methods1

MDS, myelodysplastic syndromes, VAF, variant allele frequency.

Dataset from gold standard group (n = 546) where there was agreement in local and central reviews, was used to build the model and classify genes based on maximum VAF. This model was then tested in the study population where there was disagreement in reviews (Test 1), and prospectively in a group with diagnosis subsequently confirmed by histological assessment (Test 2).


The model building process proposed seven genes as the most informative genes in predicting MDS versus Other, namely TP53, SF3B1, U2AF1, ASXL1, TET2, STAG2, and SRSF2, in order of decreasing impact. The possibility of predicting MDS was increased with increasing maximum VAF for all these genes.

In the determination of probability cutoff point, a cutoff of ≥ 0.17 was identified for predicting MDS versus Other with a sensitivity of 0.90 for true positives and specificity of 0.81 for false positives. The discrimination on the receiver operating characteristic (ROC) curve was satisfactory for the seven genes with an area under the curve (AUC) of 0.89.

The model could decrease the disagreement between local and central reviews by 74% (Table 1) and produced accurate results for the prospective cohort by 83% (Table 2). Gene modelling led to an increase in MDS diagnosis agreed both by local and central reviews compared with traditional classification (Figure 2).

Table 1. Results from Test 11

MDS, myelodysplastic syndromes; NPV, negative predictive value; PPV, positive predictive value.

Traditional classification


Gene modelling

Accuracy: 0.74


Central review


Central review

PPV: 0.89





NPV: 0.57

Local review




Local review




Sensitivity: 0.71







Specificity: 0.81

Figure 2. MDS diagnosis before and after gene modelling1

MDS, myelodysplastic syndromes.

Table 2. Results from Test 2 (prospective cohort; n = 52)1

MDS, myelodysplastic syndromes; NPV, negative predictive value; PPV, positive predictive value.


Final histopathology

Accuracy: 0.83

PPV: 0.83



NPV: 0.82

7-gene classifier




Sensitivity: 0.83




Specificity: 0.82


The use of targeted exon sequencing allowed reclassification of subjects diagnosed with MDS or Other in 74% occurrences of disagreement between local and central pathology reviews. The evaluation of an independent, prospective cohort produced results with an accuracy of 83%. Overall, study results suggest that using targeted sequencing of seven genes may improve the classification of MDS in cytopenic patients.

  1. Goll JB, Jensen TL, Lindsley RC, et al. Targeted sequencing of 7 genes can help reduce pathologic misclassification of MDS. Oral abstract #540. 62nd ASH Annual Meeting and Exposition; Dec 7, 2020; Virtual.