INFERENCE OF CELLULAR COMPOSITION FROM BULK RNA-SEQ REVEALS BIOLOGICAL HETEROGENEITY IN ACUTE MYELOID LEUKEMIA

@article{wosniaki2026inference,
  title = {{INFERENCE OF CELLULAR COMPOSITION FROM BULK RNA-SEQ REVEALS BIOLOGICAL HETEROGENEITY IN ACUTE MYELOID LEUKEMIA}},
  author = {Denise KUSMA WOSNIAKI and Bianca NICHELE KUSMA and Natalia ZURECK XAVIER and Eduarda BATISTA MENDES and Micheli DE MARCHI and Anelis Maria MARIN and Miriam PERLINGEIRO BELTRAME and Mateus NóBREGA AOKI and Dalila Lucíola ZANETTE},
  journal = {Hematology, Transfusion and Cell Therapy},
  year = {2026},
  month = oct,
  doi = {10.1016/j.htct.2026.107605},
  url = {https://doi.org/10.1016/j.htct.2026.107605},
  note = {CC BY},
}

TY  - JOUR
TI  - INFERENCE OF CELLULAR COMPOSITION FROM BULK RNA-SEQ REVEALS BIOLOGICAL HETEROGENEITY IN ACUTE MYELOID LEUKEMIA
AU  - Denise KUSMA WOSNIAKI
AU  - Bianca NICHELE KUSMA
AU  - Natalia ZURECK XAVIER
AU  - Eduarda BATISTA MENDES
AU  - Micheli DE MARCHI
AU  - Anelis Maria MARIN
AU  - Miriam PERLINGEIRO BELTRAME
AU  - Mateus NóBREGA AOKI
AU  - Dalila Lucíola ZANETTE
PY  - 2026
DA  - 2026/10/23
JO  - Hematology, Transfusion and Cell Therapy
DO  - 10.1016/j.htct.2026.107605
UR  - https://doi.org/10.1016/j.htct.2026.107605
AB  - Background: Acute myeloid leukemia (AML) is a highly heterogeneous hematologic malignancy characterized by disrupted hematopoietic differentiation and remarkable variability in morphology, molecular alterations, treatment response, and clinical outcome. Current risk stratification is primarily based on cytogenetic and molecular abnormalities, although molecular profiling has refined disease classification, the cellular architecture underlying bulk transcriptomic profiles remains poorly understood. Aim: To characterize the cellular heterogeneity of AML using BayesPrism deconvolution of bulk RNA-seq data and explore the relationship between inferred cellular composition and clinical characteristics. Methods: Bone marrow or peripheral blood samples were collected at diagnosis from patients diagnosed with AML at Erasto Gaertner Hospital following approval by the local Research Ethics Committee (CAAE 93018425.5.0000.5225). RNA sequencing was performed and sequencing reads were quality-controlled, trimmed, aligned to the GRCh38 human reference genome using the STAR aligner, and gene counts were quantified. Gene expression profiles were subsequently deconvoluted with BayesPrism using the BoneMarrowMap healthy human bone marrow single-cell RNA-seq reference atlas, enabling estimation of the relative abundance of hematopoietic cell populations. Principal component analysis (PCA) was performed to identify major patterns of cellular composition. The distribution of PCA-defined clusters was subsequently explored in relation to clinical characteristics, including European LeukemiaNet (ELN) 2022 risk classification, age, relapse, transplantation, overall survival, and APL status, through descriptive and proportional comparisons. Results: BayesPrism deconvolution identified distinct cellular composition patterns across AML samples. Principal component analysis (PCA) revealed that PC1 was predominantly driven by an inverse contribution of Monocyte (loading = +0.670) and Late GMP (loading = -0.694), separating samples into monocytic-enriched and Late GMP-enriched profiles. Secondary contributions to PC1 included Pro-Monocyte, LMPP, HSC/MPP, EoBasoMast precursor and Early GMP populations. Samples segregated into four PCA-defined clusters. The monocytic-enriched (right) cluster exhibited a higher mortality rate than the Late GMP-enriched (left) cluster (66.7% vs. 33.3%) and a higher mean age (53.4 vs. 34.6 years). At the quadrant level, Q4 showed the highest mortality (80%) and contained all ELN adverse-risk cases (3/3), whereas Q3 showed the lowest mortality (20%). Conversely, all APL cases clustered within Q2 and Q3. Overall, distinct PCA-defined cellular composition patterns coincided with differences in age, mortality, ELN risk category, and APL distribution across the cohort. These findings suggest that bulk RNA-seq deconvolution may reveal biologically meaningful variation in leukemic differentiation states that complements current molecular classification. Conclusion: Cellular deconvolution of bulk RNA-seq provides a complementary approach to characterize AML biological heterogeneity beyond conventional molecular classification. The distinct cellular composition patterns observed across AML samples suggest that deconvolution-derived cellular states may capture clinically relevant biological variation, warranting validation in larger, molecularly well-characterized cohorts.
SN  - 2531-1379
SN  - 2531-1387
ER  -