Single-cell spatial mapping reveals dynamic bone marrow microarchitectural alterations and enhances clinical diagnostics in MDS

Myelodysplastic neoplasms (MDS) are genetically heterogeneous hematopoietic malignancies characterized by ineffective hematopoiesis, peripheral cytopenias, and risk of progression to acute myeloid leukemia. Diagnosis relies on subjective histomorphologic evaluation of bone marrow biopsies, leaving the significance of subtle microarchitectural changes unclear. We applied an AI-driven, whole-slide imaging-based multiplex immunofluorescence single-cell spatial phenotyping approach to diagnostic MDS samples (n=36), serial biopsies (n=29), precursor-state samples (n=13), and normal controls (n=21), profiling over 5 million cells. Compared with age-matched controls, MDS exhibited altered progenitor cell frequencies, abnormal erythroid and megakaryocyte morphology, disrupted erythroid islands, vascular displacement of hematopoietic stem and progenitor cells, and aberrant progenitor cell clustering. Several features correlated more strongly with specific mutations (e.g., SF3B1, TP53) than with IPSS-M risk scores. We integrated these features into a composite MDS Microarchitectural Perturbation Score (MDS-MAPS) that aligned with clinical and genetic parameters across serial samples, revealing previously unrecognized, genotype-associated microarchitectural alterations in MDS. SIGNIFICANCEAI-driven spatial mapping detects previously unrecognized microarchitectural perturbations in MDS patient bone marrow biopsy tissues. Microarchitectural perturbations track closely with clinical and genetic parameters during the course of disease-modifying therapy. The detection of these features holds the potential to additionally inform clinical evaluation alongside standard clinical parameters.