AFFILIATIONS
1 Department of Chemistry, Raymond and Beverly Sackler Faculty of ExactSciences,Tel Aviv University, Tel Aviv, Israel, Department of Hematology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel
2 Department of Hematology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel, Tel Aviv, Israel
3 Department of Hematology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel, Tel Aviv, Israel
4 Department of Chemistry, Raymond and Beverly Sackler Faculty of ExactSciences,Tel Aviv University, Tel Aviv, Israel
5 Hematology, Oncology andTransfusion Medicine Center, Vilnius University Hospital Santaros Klinikos, Vilnius, Lithuania, Institute of Data Science and DigitalTechnologies, Vilnius University, Vilnius, Lithuania
6 Hematology and Oncology Department, Faculty ofMedicine, Vilnius University, Vilnius, Lithuania, Vilnius University HospitalSantaros Klinikos, Vilnius, Lithuania
7 Department of Hematology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel, Tel Aviv, Israel
8 Department of Hematology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel, Tel Aviv, Israel
9 Department of Hematology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel, Tel Aviv, Israel
10 Department of Hematology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel, Tel Aviv, Israel
11 Department of Clinical Therapeutics, National and Kapodistrian University of Athens, School of Medicine, Athens, Athens, Greece
12 Department of Hemato-Oncology,Faculty of Medicine and Dentistry, Palacky University and University Hospital,Olomouc, Olomouc, Czech Republic
13 Department of Immunology, Faculty of Medicineand Dentistry, Palacky University and University Hospital, Olomouc, Olomouc, Czech Republic
14 JaxBio Technologies Ltd., Netanya, Netanya, Israel
15 Department of Chemistry, Raymond and Beverly Sackler Faculty of ExactSciences,Tel Aviv University, Tel Aviv, Israel
ABSTRACT
Aberrant DNA methylation is well validated as a marker of disease biology and prognosis in acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS), yet epigenetic profiling remains costly and technically demanding and has not been incorporated into routine risk stratification for either disease. Here we present HemaChip, a low-cost (~$60), non-invasive epigenetic liquid-biopsy assay that detects and discriminates AML and MDS from peripheral blood on an ultra-targeted custom microarray, developed within the prospective, multicenter SANGUINE study (NCT05735704). Designed as a first-line triage step for suspected myeloid malignancy, HemaChip could spare inconclusive or unnecessary bone marrow procedures when clinical suspicion is low but present, while flagging patients who should proceed directly to bone marrow evaluation. We evaluated whether a compact biomarker panel could reproduce strong classification of AML and MDS from controls and from each other.
Peripheral blood samples from AML and MDS patients prior to treatment start and age-matched cancer-free controls were collected across four clinical centers. In a genome-wide discovery stage, bisulfite-converted genomic DNA was profiled on a high-content methylation array (~1 million loci); differential-methylation analysis compared AML vs. controls, MDS vs. controls, and AML vs. MDS identifying discriminatory regions per comparison. The selected regions, together with those from seven other SANGUINE disease cohorts, were printed on a custom Microarray featuring ~1,000 genomic loci - the HemaChip. Validation-cohort genomic DNA was fluorescently labeled and hybridized to HemaChip at participating clinical and laboratory sites; Chips were optically scanned to produce unique fluorescent pattern representing the methylation fingerprint of each clinical state. classifiers were retrained across the full chip content, selecting 70-100 discriminatory markers for each comparison, and evaluated by leave-one-out cross-validation (LOOCV) to obtain valid and unbiased performance estimates.
In total, 328 samples were analyzed to date across discovery and HemaChip, 206 healthy controls, 72 AML and 50 MDS. Discovery cohort included 48 AML, 36 MDS, and 56 controls. Genome-wide profiling separated AML from controls with 98% sensitivity and 100% specificity, and MDS from controls with 94.4% sensitivity and 100% specificity and discriminated AML from MDS with 89.6% and 97.2% sensitivity, respectively. The HemaChip validation analysis included both standard training and LOOCV analyses. 86 samples were analyzed by LOOCV, 32 AML, 22 MDS and 32 controls; in each comparison, group sizes were matched to the smaller cohort. HemaChip distinguished AML from controls with 96.9% sensitivity and 90.6% specificity (AUC 0.984), and MDS from controls with 95.5% sensitivity and specificity (AUC 0.971). AML vs. MDS yielded 72.7% and 90.9% sensitivity, respectively (AUC 0.826). Misclassifications appeared disease-driven rather than random: the single misclassified MDS case had high-risk, del(5q)-associated disease requiring hypomethylating therapy, and misclassified AML cases clustered at the AML-MDS boundary, including AML with MDS-related gene mutations or cytogenetic abnormalities, and relatively low blast-count disease. These results reflect an interim analysis: the SANGUINE myeloid arm has already recruited 266 samples, 153 AML and 113 MDS, and 324 healthy control samples. Only a subset has been run on HemaChip to date, and recruitment is ongoing.
Methylation biomarkers selected from a genome-wide discovery array and translated onto the compact, targeted HemaChip successfully distinguish AML and MDS from healthy controls and from each other, using peripheral blood, with performance that held up under LOOCV. As a low-cost, minimally invasive assay compatible with routine peripheral-blood workflows, HemaChip may help triage patients toward or away from bone marrow biopsy alongside standard clinical assessment. Accessible epigenetic data may also add an informative layer on disease biology, potentially refining risk stratification and outcome prediction in both diseases. Larger cohorts are warranted to validate this role.