Samples, xMELP, and the M-score

The xMELP assay was performed on UPENN samples as described (10, 11). Briefly, JHpaII 12XXXX primer (diluted to 3 optical density [260] units [OD] per ml, CGCCTGTTCAT) and JHpaII 24XXXX primer (6 OD/ml, CGACGTCGACTATCCATGAACAGG) were annealed by heating to 95°C for 3 minutes and slowly cooling to 25°C. Approximately 1–3 million cryopreserved cells from AML patients were thawed and washed once with PBS. Genomic DNA was extracted with Qiagen Gentra Puregene DNA isolation kit. Genomic DNA (500 ng) was then digested with either HpaII or MspI restriction enzymes and simultaneously ligated to the annealed primers by mixing genomic DNA with 7.5 μl of annealed primers, 2U T4 DNA ligase (Invitrogen), 0.5 μl ATP (100 mmol/L, pH 7.0, New England Biolabs), or either 4U of MspI (New England Biolabs) or 2U of HpaII (New England Biolabs) in a total reaction volume of 50 μl. After overnight incubation at 25°C, the reaction was diluted with 450 μl of ddH₂O, and 10 μl of the diluted DNA was subjected to PCR using 1 μl of JHpaII 24XXXX primer (6 OD/ml) and 1 μl Native Taq polymerase (ThermoFisher) in 50 μl total volume. Cycling conditions were 72°C for 10 minutes once, followed by 20 cycles of 95°C for 30 seconds and 72°C for 3 minutes, and a final 72°C extension for 10 minutes. PCR products (8 μl) were then hybridized to specific fluorescent microspheres following manufacturer’s protocol (Quantigene Plex 2.0 Assay, Affymetrix). Following hybridization, microspheres were analyzed on a FlexMAP 3D instrument running XPONENT software (Luminex). As our studies have shown, xMELP is not significantly affected by freezing samples or Ficoll centrifugation; these variables were not standardized in the assay (11).

M-score of each sample was determined using the random forest classification algorithm previously trained on an independent cohort of 344 AML samples collected by the Dutch-Belgian Hemato-Oncology Cooperative Group (HOVON) (R-scripts for M-score derivation are described and publicly available) (11). For E1900 samples, HELP-derived methylation data from BM samples were transformed to MELP-associated values using previously described regression coefficients (10).