# paper.pdf ## Metadata - PDFFormatVersion=1.6 - IsLinearized=true - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - Author=Giulia Maggioni; ; Gabriele Todisco; ; Elisabetta Sauta; ; Luca Lanino; ; Somedeb Ball; ; Jan P. Bewersdorf; ; Tariq Kewan; ; Najla H. Al Ali; ; Pierre Fenaux; ; Uwe Platzbecker; ; Valeria Santini; ; Zhuoer Xie; ; Andrew M. Brunner; ; Andrew T. Kuykendall; ; John M. Bennett; ; Rena Buckstein; ; Rafael Bejar; ; Hetty E. Carraway; ; Amy E. DeZern; ; Elizabeth A. Griffiths; ; Stephanie Halene; ; Robert P. Hasserjian; ; Alan F. List; ; Sanam Loghavi; ; Olatoyosi Odenike; ; Eric Padron; ; Mrinal M. Patnaik; ; Gail J. Roboz; ; Maximilian Stahl; ; Mikkael A. Sekeres; ; David P. Steensma; ; Michael R. Savona; ; Justin Taylor; ; Mina L. Xu; ; David A. Sallman; ; Stephen D. Nimer; ; Christopher S. Hourigan; ; Andrew H. Wei; ; Alessia Campagna; ; Marta Ubezio; ; Elena Riva; ; Denise Ventura; ; Nicole Pinocchio; ; Matteo Zampini; ; ; Alessandro Buizza; ; Antonio Russo; ; Francesco Pesce; ; Saverio D'Amico; ; Gianluca Asti; ; Mattia Delleani; ; Erica Travaglino; ; Cristina A. Tentori; ; Ivan Ferrari; ; Alessandra Crespi; ; Giulia Figini; ; Maria Di Matteo; ; Matteo Brindisi; ; Nicla Manes; ; Chiara Milanesi; ; Laura Crisafulli; ; ; Francesca Ficara; ; Gastone Castellani; ; Uma M. Borate; ; ; Fabio Efficace; ; Steven D. Gore; ; Tae K. Kim; ; ; Maria Diez‐Campelo; ; Arjan A. van de Loosdrecht; ; Navel Daver; ; Maria Rozman; ; ; Alberto Orfao; ; ; Sa A. Wang; ; Kathryn Foucar; ; Marcelo Iastrebner; ; Phillip Scheinberg; ; Yasushi Miyazaki; ; Yazan F. Madanat; ; Aref Al‐Kali; ; Moshe Mittelmann; ; ; Thomas Cluzeau; ; Lionel Ades; ; Ulrich Germing; ; Guillermo Garcia‐Manero; ; Shahram Kordasti; ; Torsten Haferlach; ; Amer M. Zeidan; ; ; ; Rami S. Komrokji; ; ; ; Matteo G. Della Porta; ; - CreationDate=D:20260609174019+05'30' - Creator=John Wiley & Sons - ModDate=D:20260609174027+05'30' - Producer=iTextSharp™ 5.5.13.4 ©2000-2024 iText Group NV (AGPL-version); modified using iText® Core 9.4.0 (AGPL version) ©2000-2025 Apryse Group NV - Subject=HemaSphere, doi:10.1002/hem3.70406 - Title=Real-world, multi-omics validation of the clinical relevance of molecular taxonomy for myelodysplastic syndromes (MDS). ° ° An international study on behalf of GenoMed4all, Synthema and icMDS consortiums - Custom.WPS-ARTICLEDOI=10.1002/hem3.70406 - Custom.WPS-JOURNALDOI=10.1002/(ISSN)2572-9241 - Custom.WPS-PROCLEVEL=3 - dc:format=application/pdf - dc:identifier=doi:10.1002/hem3.70406 - dc:title=Real-world, multi-omics validation of the clinical relevance of molecular taxonomy for myelodysplastic syndromes (MDS). ° ° An international study on behalf of GenoMed4all, Synthema and icMDS consortiums - dc:creator=Giulia Maggioni; ; Gabriele Todisco; ; Elisabetta Sauta; ; Luca Lanino; ; Somedeb Ball; ; Jan P. Bewersdorf; ; Tariq Kewan; ; Najla H. Al Ali; ; Pierre Fenaux; ; Uwe Platzbecker; ; Valeria Santini; ; Zhuoer Xie; ; Andrew M. Brunner; ; Andrew T. Kuykendall; ; John M. Bennett; ; Rena Buckstein; ; Rafael Bejar; ; Hetty E. Carraway; ; Amy E. DeZern; ; Elizabeth A. Griffiths; ; Stephanie Halene; ; Robert P. Hasserjian; ; Alan F. List; ; Sanam Loghavi; ; Olatoyosi Odenike; ; Eric Padron; ; Mrinal M. Patnaik; ; Gail J. Roboz; ; Maximilian Stahl; ; Mikkael A. Sekeres; ; David P. Steensma; ; Michael R. Savona; ; Justin Taylor; ; Mina L. Xu; ; David A. Sallman; ; Stephen D. Nimer; ; Christopher S. Hourigan; ; Andrew H. Wei; ; Alessia Campagna; ; Marta Ubezio; ; Elena Riva; ; Denise Ventura; ; Nicole Pinocchio; ; Matteo Zampini; ; ; Alessandro Buizza; ; Antonio Russo; ; Francesco Pesce; ; Saverio D'Amico; ; Gianluca Asti; ; Mattia Delleani; ; Erica Travaglino; ; Cristina A. Tentori; ; Ivan Ferrari; ; Alessandra Crespi; ; Giulia Figini; ; Maria Di Matteo; ; Matteo Brindisi; ; Nicla Manes; ; Chiara Milanesi; ; Laura Crisafulli; ; ; Francesca Ficara; ; Gastone Castellani; ; Uma M. Borate; ; ; Fabio Efficace; ; Steven D. Gore; ; Tae K. Kim; ; ; Maria Diez‐Campelo; ; Arjan A. van de Loosdrecht; ; Navel Daver; ; Maria Rozman; ; ; Alberto Orfao; ; ; Sa A. Wang; ; Kathryn Foucar; ; Marcelo Iastrebner; ; Phillip Scheinberg; ; Yasushi Miyazaki; ; Yazan F. Madanat; ; Aref Al‐Kali; ; Moshe Mittelmann; ; ; Thomas Cluzeau; ; Lionel Ades; ; Ulrich Germing; ; Guillermo Garcia‐Manero; ; Shahram Kordasti; ; Torsten Haferlach; ; Amer M. Zeidan; ; ; ; Rami S. Komrokji; ; ; ; Matteo G. Della Porta; ;,Giulia Maggioni,Gabriele Todisco,Elisabetta Sauta,Luca Lanino,Somedeb Ball,Jan P. Bewersdorf,Tariq Kewan,Najla H. Al Ali,Pierre Fenaux,Uwe Platzbecker,Valeria Santini,Zhuoer Xie,Andrew M. Brunner,Andrew T. Kuykendall,John M. Bennett,Rena Buckstein,Rafael Bejar,Hetty E. Carraway,Amy E. DeZern,Elizabeth A. Griffiths,Stephanie Halene,Robert P. Hasserjian,Alan F. List,Sanam Loghavi,Olatoyosi Odenike,Eric Padron,Mrinal M. Patnaik,Gail J. Roboz,Maximilian Stahl,Mikkael A. Sekeres,David P. Steensma,Michael R. Savona,Justin Taylor,Mina L. Xu,David A. Sallman,Stephen D. Nimer,Christopher S. Hourigan,Andrew H. Wei,Alessia Campagna,Marta Ubezio,Elena Riva,Denise Ventura,Nicole Pinocchio,Matteo Zampini,Alessandro Buizza,Antonio Russo,Francesco Pesce,Saverio D'Amico,Gianluca Asti,Mattia Delleani,Erica Travaglino,Cristina A. Tentori,Ivan Ferrari,Alessandra Crespi,Giulia Figini,Maria Di Matteo,Matteo Brindisi,Nicla Manes,Chiara Milanesi,Laura Crisafulli,Francesca Ficara,Gastone Castellani,Uma M. Borate,Fabio Efficace,Steven D. Gore,Tae K. Kim,Maria Diez‐Campelo,Arjan A. van de Loosdrecht,Navel Daver,Maria Rozman,Alberto Orfao,Sa A. Wang,Kathryn Foucar,Marcelo Iastrebner,Phillip Scheinberg,Yasushi Miyazaki,Yazan F. Madanat,Aref Al‐Kali,Moshe Mittelmann,Thomas Cluzeau,Lionel Ades,Ulrich Germing,Guillermo Garcia‐Manero,Shahram Kordasti,Torsten Haferlach,Amer M. Zeidan,Rami S. Komrokji,Matteo G. Della Porta - dc:publisher=John Wiley & Sons, Ltd - dc:description=HemaSphere, doi:10.1002/hem3.70406 - prism:doi=10.1002/hem3.70406 - prism:url=http://dx.doi.org/10.1002/hem3.70406 - prism:aggregationtype=journal - prism:publicationname=HemaSphere - prism:issn=2572-9241 - prism:volume=10 - prism:number=6 - jav:journal_article_version=VoR - xmp:creatortool=John Wiley & Sons - xmp:modifydate=2026-06-09T17:40:27+05:30 - xmp:createdate=2026-06-09T17:40:19+05:30 - xmp:metadatadate=2026-06-09T17:40:27+05:30 - xmprights:marked=True - pdf:producer=iTextSharp™ 5.5.13.4 ©2000-2024 iText Group NV (AGPL-version); modified using iText® Core 9.4.0 (AGPL version) ©2000-2025 Apryse Group NV - pdfx:wps-articledoi=10.1002/hem3.70406 - pdfx:wps-journaldoi=10.1002/(ISSN)2572-9241 - pdfx:wps-proclevel=3 - xmpmm:documentid=uuid:fd375657-975a-428e-a813-f2660371e9a7 - xmpmm:instanceid=uuid:ba941e9a-6fa6-441b-badc-411132769945 - pdfuaid:part=1 - pdfaextension:schemas=http://ns.adobe.com/pdf/1.3/pdfAdobe PDF SchemainternalA name object indicating whether the document has been modified to include trapping informationTrappedTexthttp://ns.adobe.com/xap/1.0/mm/xmpMMXMP Media Management SchemainternalUUID based identifier for specific incarnation of a documentInstanceIDURIinternalThe common identifier for all versions and renditions of a document.OriginalDocumentIDURIhttp://www.aiim.org/pdfua/ns/id/pdfuaidPDF/UA ID SchemainternalPart of PDF/UA standardpartInteger ## Contents ### Page 1 Received: 21 October 2025 | Accepted: 6 May 2026 DOI: 10.1002/hem3.70406A R T I C L E Real‐world, multi‐omics validation of the clinical relevance of molecular taxonomy for myelodysplastic syndromes (MDS) Giulia Maggioni1,^ | Gabriele Todisco1,2,^ | Elisabetta Sauta1,^ | Luca Lanino3 |Somedeb Ball4 | Jan P. Bewersdorf3 | Tariq Kewan3 | Najla H. Al Ali5 |Pierre Fenaux6 | Uwe Platzbecker7 | Valeria Santini8 | Zhuoer Xie5 |Andrew M. Brunner9 | Andrew T. Kuykendall5 | John M. Bennett10 |Rena Buckstein11 | Rafael Bejar12 | Hetty E. Carraway13 | Amy E. DeZern14 |Elizabeth A. Griffiths15 | Stephanie Halene3 | Robert P. Hasserjian16 |Alan F. List17 | Sanam Loghavi18 | Olatoyosi Odenike19 | Eric Padron5 |Mrinal M. Patnaik20 | Gail J. Roboz21 | Maximilian Stahl3 | Mikkael A. Sekeres22 |David P. Steensma23 | Michael R. Savona4 | Justin Taylor22 | Mina L. Xu24 |David A. Sallman5 | Stephen D. Nimer22 | Christopher S. Hourigan25 |Andrew H. Wei26 | Alessia Campagna1 | Marta Ubezio1 | Elena Riva1 |Denise Ventura1 | Nicole Pinocchio1 | Matteo Zampini1 |Alessandro Buizza1,2 | Antonio Russo1 | Francesco Pesce2 | Saverio D'Amico1 |Gianluca Asti1 | Mattia Delleani1 | Erica Travaglino27 | Cristina A. Tentori1 |Ivan Ferrari1,2 | Alessandra Crespi1,2 | Giulia Figini1,2 | Maria Di Matteo1,2 |Matteo Brindisi1 | Nicla Manes1 | Chiara Milanesi1 | Laura Crisafulli1 |Francesca Ficara1 | Gastone Castellani28 | Uma M. Borate29 |Fabio Efficace30 | Steven D. Gore31 | Tae K. Kim4 | Maria Diez‐Campelo32 |Arjan A. van de Loosdrecht33 | Navel Daver34 | Maria Rozman35 |Alberto Orfao32 | Sa A. Wang18 | Kathryn Foucar36 | Marcelo Iastrebner37 |Phillip Scheinberg38 | Yasushi Miyazaki39 | Yazan F. Madanat40 | Aref Al‐Kali20 |Moshe Mittelmann41 | Thomas Cluzeau42 | Lionel Ades6 | Ulrich Germing43 |Guillermo Garcia‐Manero34 | Shahram Kordasti44 | Torsten Haferlach45 |Amer M. Zeidan3,^^ | Rami S. Komrokji5,^^ | Matteo G. Della Porta1,2,^^ | An international study on behalf of GenoMed4all, Synthema, and icMDS consortia ### Page 2 Graphical Abstract ### Page 3 Received: 21 October 2025 | Accepted: 6 May 2026 DOI: 10.1002/hem3.70406A R T I C L E Real‐world, multi‐omics validation of the clinical relevance of molecular taxonomy for myelodysplastic syndromes (MDS) Giulia Maggioni1,^ | Gabriele Todisco1,2,^ | Elisabetta Sauta1,^ | Luca Lanino3 |Somedeb Ball4 | Jan P. Bewersdorf3 | Tariq Kewan3 | Najla H. Al Ali5 |Pierre Fenaux6 | Uwe Platzbecker7 | Valeria Santini8 | Zhuoer Xie5 |Andrew M. Brunner9 | Andrew T. Kuykendall5 | John M. Bennett10 |Rena Buckstein11 | Rafael Bejar12 | Hetty E. Carraway13 | Amy E. DeZern14 |Elizabeth A. Griffiths15 | Stephanie Halene3 | Robert P. Hasserjian16 |Alan F. List17 | Sanam Loghavi18 | Olatoyosi Odenike19 | Eric Padron5 |Mrinal M. Patnaik20 | Gail J. Roboz21 | Maximilian Stahl3 | Mikkael A. Sekeres22 |David P. Steensma23 | Michael R. Savona4 | Justin Taylor22 | Mina L. Xu24 |David A. Sallman5 | Stephen D. Nimer22 | Christopher S. Hourigan25 |Andrew H. Wei26 | Alessia Campagna1 | Marta Ubezio1 | Elena Riva1 |Denise Ventura1 | Nicole Pinocchio1 | Matteo Zampini1 |Alessandro Buizza1,2 | Antonio Russo1 | Francesco Pesce2 | Saverio D'Amico1 |Gianluca Asti1 | Mattia Delleani1 | Erica Travaglino27 | Cristina A. Tentori1 |Ivan Ferrari1,2 | Alessandra Crespi1,2 | Giulia Figini1,2 | Maria Di Matteo1,2 |Matteo Brindisi1 | Nicla Manes1 | Chiara Milanesi1 | Laura Crisafulli1 |Francesca Ficara1 | Gastone Castellani28 | Uma M. Borate29 |Fabio Efficace30 | Steven D. Gore31 | Tae K. Kim4 | Maria Diez‐Campelo32 |Arjan A. van de Loosdrecht33 | Navel Daver34 | Maria Rozman35 |Alberto Orfao32 | Sa A. Wang18 | Kathryn Foucar36 | Marcelo Iastrebner37 |Phillip Scheinberg38 | Yasushi Miyazaki39 | Yazan F. Madanat40 | Aref Al‐Kali20 |Moshe Mittelmann41 | Thomas Cluzeau42 | Lionel Ades6 | Ulrich Germing43 |Guillermo Garcia‐Manero34 | Shahram Kordasti44 | Torsten Haferlach45 |Amer M. Zeidan3,^^ | Rami S. Komrokji5,^^ | Matteo G. Della Porta1,2,^^ | An international study on behalf of GenoMed4all, Synthema, and icMDS consortia Correspondence: Amer M. Zeidan (amer.zeidan@yale.edu); Rami S. Komrokji (Rami.Komrokji@moffitt.org) and Matteo G. Della Porta (matteo.della_porta@ hunimed.eu)HemaSphere. 2026;10:e70406.3 of 17 | hemaspherejournal.com https://doi.org/10.1002/hem3.70406This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made. © 2026 The Author(s). HemaSphere published by John Wiley & Sons Ltd on behalf of European Hematology Association. ### Page 4 AbstractMyelodysplastic syndromes (MDS) are clinically and biologically diverse disorders, emphasizing the need for personalized treatment approaches. The International Working Group for Prognostication of MDS (IWG_PM) recently introduced a molecular classification, referred to as the MDS taxonomy, that categorizes patients into 16 subgroups based on 21 gene mutations, 6 cytogenetic abnormalities, and loss of heterozygosity (LOH) at TP53 and TET2 loci. This study sought to validate and enhance the clinical relevance of the MDS taxonomy by analyzing a large retrospective cohort (n = 5136) and transcriptomic data from a prospective cohort (n = 477). The taxonomy successfully identified subgroups with distinct clinical characteristics and disease progression patterns. However, incorporating gene interactions from taxonomy subgroups did not improve the prognostic performance of the Molecular International Prognostic Scoring System (IPSS‐M). We further assessed whether the taxonomy could guide management in patients receiving disease‐modifying therapies. Except for the “TP53‐complex” subgroup, taxonomy classifications were not predictive of hypomethylating agent response or transplant outcomes. Nonetheless, they correlated with overall survival, suggesting that while both IPSS‐M and the taxonomy capture disease biology, other non‐genetic factors may influence treatment response. RNA sequencing confirmed the biological distinctiveness of the taxonomy groups. Transcriptomic profiling of CD34+ bone marrow cells revealed unique, homogeneous gene expression patterns, particularly within the AML‐like, biTET2, SF3B1, and TP53‐complex subgroups. Further integration of multi‐omics data may refine MDS classification, improving clinical decision‐making and guiding the development of targeted therapies.INTRODUCTIONMyelodysplastic syndromes/neoplasms (MDS) are characterized by ineffective hematopoiesis and increased risk of evolution into acute myeloid leukemia (AML).1,2 Over the years, the World Health Orga- nization (WHO) classification of myeloid neoplasms, primarily based on morphology has served as a status for diagnosis and classification, guiding treatment decisions, clinical trial eligibility, and drug approvals in patients with MDS.3–5 Recent advancements in genome characterization have re- volutionized the study of MDS.2 High proportions of patients exhibit recurrent somatic mutations that drive clinical phenotype and disease evolution on an individual level.6.7 Accordingly, there is a shift from traditional classification schemes, primarily based on morphological 1IRCCS Humanitas Research Hospital, Rozzano, Milan, Italy 2Department of Biomedical Sciences, Humanitas University, Pieve Emanuele, Milan, Italy 3Hematology Section, Department of Internal Medicine, Yale School of Medicine, New Haven, USA 4Vanderbilt Ingram Cancer Center and Vanderbilt University School of Medicine, Nashville, Tennessee, USA 5Department of Malignant Hematology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, Florida, USA 6Hôpital Saint‐Louis, Université Paris Cité, Paris, France 7Universitätsklinikum Carl Gustav Carus, Dresden, Germany 8University of Florence, Florence, Italy 9Leukemia Program, Massachusetts General Hospital Cancer Center, Harvard Medical School, Boston, Massachusetts, USA 10University of Rochester Medical Center, Rochester, New York, USA 11Odette Cancer Centre, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada 12University of California San Diego, La Jolla, California, USA 13Leukemia Program, Taussig Cancer Institute, Cleveland Clinic, Cleveland, Ohio, USA 14The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University, Baltimore, Maryland, USA 15Roswell Park Comprehensive Cancer Center, Buffalo, New York, USA 16Massachusetts General Hospital, Boston, Massachusetts, USA 17Precision BioSciences, Durham, North Carolina, USA 18Department of Hematopathology, The University of Texas MD Anderson Cancer Center, Houston, Texas, USA 19University of Chicago Medicine and University of Chicago Comprehensive Cancer Center, Chicago, Illinois, USA 20Department of Internal Medicine, Division of Hematology, Mayo Clinic, Rochester, Minnesota, USA 21New York‐Presbyterian/Weill Cornell Medicine, New York, New York, USA 22Sylvester Comprehensive Cancer Center, University of Miami, Miami, Florida, USA 23Ajax Therapeutics, Cambridge, Massachusetts, USA 24Department of Pathology & Laboratory Medicine, Yale School of Medicine, New Haven, Connecticut, USA 25Virginia Tech FBRI Cancer Research Center, Washington, District of Columbia, USA 26Department of Haematology, Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia 27IRCCS Policlinico San Matteo, Pavia, Italy 28Department of Medical and Surgical Sciences—DIMEC, University of Bologna, Bologna, Italy 29Department of Internal Medicine, Division of Hematology, The Ohio State University, Columbus, Ohio, USA 30Italian Group for Adult Haematologic Diseases (GIMEMA), Rome, Italy 31Division of Cancer Prevention, National Cancer Institute, Rockville, Maryland, USA 32Department of Hematology, Hospital Universitario de Salamanca, Salamanca, Spain 33Department of Hematology, VU University Medical Center—Cancer Center, Amsterdam, Netherlands 34Department of Leukemia, The University of Texas MD Anderson Cancer Center, Houston, Texas, USA 35Hospital Clínic Barcelona, Barcelona, Spain 36University of New Mexico, Albuquerque, New Mexico, USA 37Sanatorio Sagrado Corazón, Buenos Aires, Argentina 38Hospital A Beneficência Portuguesa de São Paulo, Sao Paulo, Brazil 39Atomic Bomb Disease Institute, Nagasaki University, Nagasaki, Japan 40Division of Hematology and Oncology, UT Southwestern, Dallas, Texas, USA 41Tel Aviv Sourasky Medical Center, Sackler Faculty of Medicine, Tel Aviv University, Tel‐Aviv, Israel 42Nice University Hospital, Nice, France 43Department of Hematology, Oncology, and Clinical Immunology, Heinrich‐Heine‐University, Düsseldorf, Germany 44Department of Haematology, Guy's Hospital & Comprehensive Cancer Centre, King′s College, London, UK 45MLL Munich Leukemia Laboratory, Munich, Germany ^Giulia Maggioni, Gabriele Todisco, and Elisabetta Sauta contributed equally as first authors. ^^Amer M. Zeidan, Rami S. Komrokji, and Matteo G. Della Porta contributed equally as senior authors. HemaSphere | 4 of 17 ### Page 5 and clinical features, to next‐generation systems that integrate genomic data, providing a more accurate representation of the dis- ease biology and better capturing clinical‐pathological entities.6–8 In 2022, two distinct classification systems for MDS were re- leased by the WHO and the International Consensus Classification (ICC) committees, incorporating new molecular information to iden- tify specific entities.9,10 In addition to the entity of MDS with del 5q, WHO and ICC systems introduced new genetically‐defined cate- gories based on SF3B1 and TP53 mutations.9–11 Both classifications still included morphologically‐defined categories which, however, fail to identify groups of patients with distinct genomic profiles.8,12 To expand the integration of molecular features into the classifi- cation of MDS, the International Working Group for Prognostication of MDS (IWG‐PM) recently proposed a new molecular taxonomy based on 16 groups, obtained using information from 21 genes, 6 cytogenetic events, and loss of heterozygosity (LOH) at the TP53 and TET2 loci.13 The groups varied in size from 0.5% to 14% of patients and were associated with distinct clinical phenotypes and outcomes. The authors proposed that this MDS taxonomy could serve as a basis for future classification schemas and translational therapeutic research.13 In this project, leveraging a large retrospective real‐world popu- lation of MDS patients with genomic annotation along with treatment and outcome data, as well as transcriptomic information from a prospective patient cohort, we aimed to validate and expand the clinical relevance of MDS taxonomy by: (i) evaluating its predictive and prognostic value in patients receiving disease‐modifying treat- ments, including hypomethylating agents (HMA) and allogeneic stem cell transplantation (HSCT); (ii) assessing the biological homogeneity within genetically‐defined entities by analyzing the RNA‐sequencing profiles of isolated CD34+ bone marrow progenitor cells.MATERI ALS A ND METHODS Study populationsThe study was conducted by GenoMed4All, Synthema, and the In- ternational Consortium for MDS (icMDS), with the support of Euro- BloodNET. The Humanitas Ethics Committee approved the study (ClinicalTrials.gov Identifier: NCT04889729). Informed consent was obtained from all participants. The study population consisted of two patient cohorts with available genomic data, allowing for stratification according to the MDS taxonomy. The first cohort comprised an international retrospective dataset of 5136 patients. Inclusion criteria were age >18 years, a diagnosis of primary or therapy‐related MDS according to the 2016 WHO classifi- cation,3 and availability of IPSS‐M–related variables14 collected at di- agnosis and prior to the initiation of any disease‐modifying treatment. The second cohort consisted of 447 patients prospectively enrolled at Humanitas Research Hospital (Milan, Italy). Eligible patients had a di- agnosis of either primary MDS or AML evolving from MDS, according to the 2016 WHO classification criteria.3 In addition to clinical and genomic data, transcriptomic data from CD34+ bone marrow progenitors col- lected at diagnosis were available for this cohort. All cases were subsequently reclassified according to the 2022 WHO and ICC criteria and then stratified using both IPSS‐R and IPSS‐M risk categories.14,15DNA/RNA sequencingMutation screening of genes related to myeloid neoplasms was per- formed on DNA from peripheral blood leucocytes or bone marrow mononuclear cells.6,16 Oncogenic mutations with a minimum variant allele frequency (VAF) of 2% were considered. All patients included in the study had the necessary mutational data to allow proper classi- fication according to the MDS taxonomy framework. Information on copy‐neutral loss of heterozygosity (cnLOH) events was not available; therefore, biallelic inactivation of TET2 and TP53 genes was inferred based on mutation number, VAF, and conventional karyotyping data. RNA was extracted from CD34+ bone marrow progenitors pur- ified by MACS microbead kit (Miltenyi Biotech). RNA was isolated through RNeasy Micro kit (Qiagen); quality control was performed with the Agilent 2200 TapeStation system (Agilent), and RNA with RIN > 7 were used for library preparation. Libraries were prepared by using the SMART‐Seq v4 Ultra Low Input RNA Kit (Takara). Libraries were sequenced using the Novaseq. 6000 (Illumina) with paired‐end 150‐bp configuration and aligned against hg38.17 Uniquely mapped read pairs were counted by featureCounts and used for gene ex- pression analysis.18–20 Differential expression analysis was performed with DESeq. 2.21 Genes with raw read counts <50 in less than 10% of samples were filtered out. The significance of differential gene ex- pression among groups was calculated using likelihood ratio test in- cluding sequencing batch and blast count in the reduced model. Genes with false discovery rate (FDR) < 0.0001 were considered significantly differentially expressed. Raw read counts were normal- ized by DESeq. 2 by applying variance stabilizing transformation and used for the graphical representation after logarithmic transforma- tion. Pair‐wise Pearson correlation coefficients were calculated for all samples using differentially expressed normalized gene expression data. Hierarchical clustering was performed using the calculated Pearson correlation coefficients as the similarity matrix, using the average linkage method. To further explore the relationship between transcriptome and genotype, a random forest‐based classification was carried out using caret R package22 to predict MDS taxonomy from differentially expressed genes after splitting the cohort in training (0.7) and validation (0.3) set (mtry = 34, cross fold number = 10, repeats = 3). ClusterProfiler package was used for gene pathway enrichment analyses.23Statistical analysisNumerical variables were summarized by median and range; catego- rical variables were described with count and relative frequency (%) of subjects in each category. Survival curves were estimated with the Kaplan–Meier method and differences among groups were evaluated by log‐rank test. Overall survival (OS) and leukemia‐free survival (LFS) were defined as the time between diagnosis and death (from any cause) or last follow‐up (for censored observations) and the time between diagnosis and AML evolution (if any) or last follow‐up (for censored observations), respectively. Cumulative risk of AML evolution was modeled using Fine‐Gray regression. When focusing on patient populations receiv- ing a specific treatment, OS was calculated as the time between the start of treatment and death/last follow‐up. The probability of relapse after treatment was estimated considering treatment as a failure at the time of hematologic relapse according to standardized criteria.24 Multivariable survival analyses were performed by Cox's pro- portional hazards regression models. We investigated the prognostic effect of MDS taxonomy co‐mutation patterns using multivariate proportional hazards Cox's models with interaction terms across the considered clinical endpoints. The discriminatory power of the pre- dictive models and the relative goodness of fit were evaluated using Harrell's concordance index (C‐index).25 All the analyses were performed using R software (4.4.1 version).5 of 17 | ### Page 6 T A B L E 1 Characteristics of study populations.Retrospective cohort (n = 5136) Prospective cohort (n = 447)Demographics Age (years) 69 (18–97) 68 (21–93) Sex Male 3199 (62.3%) 278 (62.2%) Female 1937 (37.7%) 169 (37.8%) Clinical variables Absolute neutrophil count (103/μL) 1.6 (0.0–29.0) 1.5 (0.0–19.7) Hemoglobin concentration (g/dL) 9.6 (2.1–16.2) 9.5 (4.4–15.8) Platelet count (×103/μL) 107 (2–680) 107 (1–440) WHO 2022 classification MDS del(5q) 216 (4.2%) 14 (3.1%) MDS mutated SF3B1 621 (12.1%) 59 (13.2%) MDS biallelic TP53 inactivation 655 (12.8%) 12 (2.7%) MDS low blasts 1497 (29.1%) 125 (28%) MDS low blasts with ring sideroblasts 154 (3.0%) 20 (4.5%) MDS hypoplastic 94 (1.8%) 0 (0%) MDS increased blasts 1 797 (15.5%) 36 (8.1%) MDS increased blasts 2 936 (18.2%) 37 (8.3%) MDS with fibrosis 114 (2.2%) 1 (0.2%) AML with NPM1 mutation 30 (0.6%) 43 (9.6%) AML with other defining genetic abnormalities 22 (0.4%) 0 (0%) AML, defined by differentiation 0 (0%) 15 (3.4%) AML, MDS‐related 0 (0%) 85 (19.0%) ICC 2022 classification MDS del(5q) 220 (4.3%) 14 (3.1%) MDS mutated SF3B1 516 (10.0%) 57 (12.8%) MDS mutated TP53 484 (9.4%) 11 (2.5%) MDS, NOS w/o dysplasia 31 (0.6%) 2 (0.4%) MDS single‐lineage dysplasia 465 (9.0%) 52 (11.6%) MDS multilineage dysplasia 1339 (26.1%) 94 (21.0%) MDS excess blasts 843 (16.4%) 35 (7.8%) MDS–AML mutatedTP53260 (5.0%) 6 (1.3%) MDS–AML gene mutations 559 (10.9%) 22 (4.9%) MDS–AML cytogenetic abnormalities 113 (2.2%) 3 (0.7%) MDS–AML not otherwise specified 261 (5.1%) 9 (2.0%) AML with mutated NPM1 13 (0.3%) 42 (9.4%) AML with mutated TP53 0 (0%) 27 (6.0%) AML with recurrent genetic abnormalities 31 (0.6%) 0 (0.0%)T A B L E 1 (Continued)Retrospective cohort (n = 5136) Prospective cohort (n = 447)AML‐MDS‐related gene mutations 0 (0%) 43 (9.6%) AML‐MDS‐related cytogenetic abnormalities 0 (0%) 19 (4.3%) AML not otherwise specified 1 (0.1%) 11 (2.5%)IPSS‐R (n = 5084) (n = 304)Very Low 574 (11.3%) 76 (25.0%) Low 1441 (28.3%) 109 (35.9%) Intermediate 946 (18.6%) 56 (18.4%) High 768 (15.1%) 35 (11.5%) Very High 876 (17.2%) 27 (8.9%) Not Available 479 (9.4%) 1 (0.3%)IPSS‐M (n = 5084) (n = 304)Very Low 329 (6.5%) 47 (15.5%) Low 1410 (27.7%) 125 (41.1%) Moderate Low 670 (13.2%) 42 (13.8%) Moderate High 615 (12.1%) 22 (7.2%) High 994 (19.6%) 33 (10.9%) Very High 1066 (20.1%) 25 (11.5%) MDS taxonomy classificationDDX41 96 (1.9%) 0 (0%) AML‐like 81 (1.6%) 43 (9.6%)TP53‐complex 1031 (20.1%) 84 (18.8%) t(1;7) 24 (0.5%) 0 (0%)‐7/SETBP1 261 (5.1%) 17 (3.8%) del(5q) 202 (3.9%) 18 (4.0%)EZH2‐ASXL1 877 (17.1%) 8 (1.8%)IDH‐STAG2 173 (3.4%) 46 (10.3%)BCOR/L1 113 (2.2%) 0 (0%) bi‐TET2 335 (6.5%) 37 (8.3%)U2AF1157 56 (1.1%) 32 (7.2%)U2AF134 93 (1.8%) 0 (0%)SRSF2 98 (1.9%) 0 (0%)ZRSR2 65 (1.3%) 8 (1.8%)SF3B1 480 (9.3%) 54 (12.1%) CCUS‐like 301 (5.9%) 35 (7.8%) mNOS 408 (7.9%) 32 (7.2%) No‐event 442 (8.6%) 26 (5.8%) Disease‐modifying treatments Hypomethylating Agents (HMA) 1915/5136 (37.3%) 120/447 (26.8%) Stem cell transplantation (HSCT) 1317/5136 (25.9%) 71/447 (15.9%)Note: Data are median (range) or n (%). Abbreviations: IPSS‐M, Molecular International Prognostic Scoring System; IPSS‐R, Revised International Prognostic Scoring System; MDS, myelodysplastic syndrome. HemaSphere | 6 of 17 ### Page 7 RESULTS Clinical features of MDS patients stratified into taxonomy molecular subgroupsClinical features of 5136 patients at diagnosis were reported in Table 1. Frequency of recurrent gene mutations and chromosomal abnormalities was reported in Supporting Information Appendix (Figure S1). Date range of diagnosis was from 1990 to 2021. Median age at diagnosis was 67 years, range 18–97. Follow‐up was updated on 02/2022. We applied hierarchical MDS taxonomy rules to reclassify the study population,13 resulting in 18 subgroups with a distribution percentage akin to the original IWG‐PM cohort. The 16 molecular groups encompassed 83.5% (4286) of patients, relying on the pre- sence or absence of mutations in 21 genes (ASXL1, BCOR, BCORL1, DDX41, DNMT3A, EZH2, FLT3, IDH1, IDH2, MLL, MYC, NPM1, SETBP1, SF3B1, SRSF2, STAG2, TET2, TP53, U2AF1, WT1, ZRSR2), 6 cytogenetic events (inv(3), complex karyotype, der(1;7), −7, del(5q),‐Y), and LOH at the TP53 and TET2 loci. The molecularly not‐otherwise specified (mNOS) group (7.9%, n = 408) was char- acterized by the presence of other cytogenetic abnormalities and/or mutations in 29 additional recurrently mutated genes (>0.5%). The“No‐event” category (8.6%, n = 442) was defined by the absence of any recurrent drivers assessed with our technique and aligns with the percentage of patients reported in the original article (Table 1 and Supporting Information S1: Appendix Table S1). The redistribution of(A)F I G U R E 1 (A) Association between MDS Taxonomy molecular groups and clinical phenotypes in the retrospective cohort (n = 5136). Colors towards brown indicate a mutually exclusive relationship, while colors towards dark green indicate co‐occurrence. P values are from Fisher Exact test based on 2 × 2 contingency tables comparing the presence versus absence of each clinical phenotype in each molecular group versus all other patients, with Benjamini–Hochberg multiple testing correction (False Discovery Rate, FDR). (B) MDS Taxonomy groups characterization. For each molecular category, the relative proportions of Bone Marrow (BM) blasts (upper left), IPSS‐M risk categories (upper right), the number of cytogenetic alterations per patient (lower left), and the number of mutated genes per patient (lower right) are indicated as stacked bar plots. (C, D) Association between MDS Taxonomy molecular groups and clinical outcomes, including overall survival (C) and cumulative risk of AML transformation (D). Diamonds represent the median OS (C) and AML transformation risk (D), with horizontal blue bars indicating the 95% confidence intervals. Global differences among molecular groups were assessed using the Log‐Rank test (for OS) and the Gray's test (for AML transformation). Hazard ratios (HRs) with 95% confidence intervals reported for each group were derived from Cox proportional hazards model for OS and from Fine–Gray competing risk model for AML transformation, using the “No‐event” group as reference. AML, acute myeloid leukemia; MDS, myelodysplastic syndromes/neoplasms; OS, overall survival.7 of 17 | ### Page 8 patients across different classification systems is shown in the Sup- porting Information S1: Appendix Figure S6. The groups exhibited variability in both size (ranging from 0.5% to 20.1% of the cohort) and molecular complexity (with a median of 0 to 5 mutated genes per patient). They demonstrated associations with distinct demographic and clinical presentations (Age, sex, per- ipheral blood counts and transfusion dependence, Figure 1A), mor- phological, genetic characteristics and prognostication (percentage of bone marrow blast, cytogenetic risk according to IPSS‐R criteria15 and IPSS‐M risk categories14 Figure 1B), and patient outcomes (Overall survival and risk of leukemic evolution, Figure 1C,D). Detailed information on distinct molecular subgroups (including therapy‐related MDS) and the analysis of the prognostic influence of bone marrow blasts within genetic subtypes, together with the fre- quency of gene mutations and chromosomal alterations and patterns of co‐occurrence and mutual‐exclusivity of molecular alterations for each molecular group were reported in Supporting Information S1: Appendix Figures S2 and S3.Prognostic effect of MDS taxonomyWe aimed to explore the potential added prognostic value of MDS taxonomy with respect to conventional stratification according to IPSS‐R and IPSS‐M criteria. We therefore calculated IPSS‐R and IPSS‐M in the study cohort at diagnosis.14,15 Cytogenetic abnormalities were classified according to the IPSS‐R criteria.15 For IPSS‐M calculation, gene mutations were considered as binary variables with the exception of TP53 (not mu- tated, mono‐allelic mutation, multi‐hit mutations) and SF3B1(SF3B15q [SF3B1 mutation in the presence of isolated del(5q) only or with one additional aberration excluding −7/del(7q)], and SF3B1α[SF3B1 mutation without co‐mutations in BCOR, BCORL1, RUNX1, NRAS, STAG2, SRSF2, and del(5q)]14 (Table 1). We analyzed the probability of OS and risk of AML evolution for all MDS taxonomy subgroups (Figure 1C,D). In this analysis, patients who received HSCT were censored at the time of the procedure. MDS taxonomy categories showed significantly different probabilitiesF I G U R E 1 (Continued) HemaSphere | 8 of 17 ### Page 9 (C) (D)F I G U R E 1 (Continued)9 of 17 | ### Page 10 of both OS and risk of AML evolution (ranging from 1.3 to 10.9 years P < 0.001, and from 0 to 19.1%, P < 0.001). The independent effect of MDS taxonomy on clinical outcome was maintained in a multivariable model including age, sex, neutrophil, hemoglobin, and platelet count as covariates (HR 1.37 [1.29–1.44], P < 0.001 for OS and HR 1.31 [1.22–1.39], P < 0.001 for risk of AML evolution). Analysis of overall survival without censoring at transplantation confirmed that the taxonomy retained its prognostic significance and that its perfor- mance is robust and not driven by transplant‐related censoring (Supporting Information S1: Appendix Figure S5). The MDS taxonomy did not demonstrate superior performance compared to the conventional IPSS‐R and IPSS‐M scoring systems. Regarding OS, the concordance index (CI) for MDS taxonomy was 0.70 (0.69–0.72) versus 0.72 (0.71–0.73) and 0.78 (0.77–0.80) for IPSS‐R and IPSS‐M, respectively Considering leukemia‐free survival (LFS), the CI for MDS taxonomy was 0.69 (0.67–0.71) compared to 0.74 (0.72–0.75) and 0.83 (0.82–0.85) for IPSS‐R and IPSS‐M, respectively (Supporting Information S1: Appendix Figure S4). Additionally, IPSS‐M retained a significant effect on the prognostic stratification of MDS taxonomy groups (HR 1.47 [1.41–1.54] for OS and 1.67 [1.49–1.53] for LFS). Next, we assessed whether incorporating information on gene interactions (as defined by the MDS taxonomy) could improve the prognostic performance of the IPSS‐M score. To this end, we devel- oped two multivariable Cox models: one including IPSS‐M features alone and another incorporating gene interactions as model's covari- ates. These interactions terms were selected based on the interaction patterns described in the MDS taxonomy and observed in our patient's cohort. The inclusion of gene interactions did not result in a meaningful improvement in the CI of IPSS‐M, with only marginal increases of 0.01 for OS and 0.002 for LFS, compared to the reference model.Predictive and prognostic effect of MDS taxonomy in patients receiving specific treatmentsIn the context of MDS, a growing number of patients undergo disease‐modifying therapies, such as Hematopoietic Stem Cell Transplantation (HSCT) and Hypomethylating Agents (HMA, parti- cularly for high‐risk individuals ineligible for HSCT).4,5 Consequently, we aimed to evaluate whether MDS taxonomy can discern differ- ences in the likelihood of response to specific treatments (predictive value) and effectively stratify the probability of post‐treatment sur- vival (prognostic value) (Figure 2). To explore the predictive and prognostic value of these subgroups, we conducted analyses in two distinct populations from the cohort, one receiving HSCT and the other undergoing HMA treatment. In evaluating the predictive value of the classification, primary endpoints included the overall response rate (comprising complete response (CR), partial re- sponse (PR), marrow CR (mCR), and Stable Disease with Hematological Improvement (SD‐HI) based on the 2006 International Working Group (IWG) criteria) for patients treated with HMA, and the risk of disease relapse for those treated with HSCT.24 Simultaneously, the prognostic value of these subgroups was examined based on the probability of OS calculated from the starting date of treatment in both cases. We first assessed the predictive and prognostic impact of MDS taxonomy in a cohort of high‐risk MDS patients who received HMA according to the current therapy guidelines.4,5 A total of 1915 sub- jects were included in the analysis. The median duration of MDS prior to HMA initiation was 6.5 months (range: 1–119). Patients received a median of 4 HMA cycles (range: 1–21), with no significant differences observed across molecular subgroups (P = 0.96). The overall response rate, assessed after 4–6 treatment cycles, was 30.7%, with no significant variation among the different MDS taxonomy categories (P = 0.85). In the HMA‐treated population, the median OS following HMA initiation was 13.8 months (range: 1–119). MDS taxonomy had a significant impact on post‐HMA survival outcomes (P < 0.001). Patients in theSF3B1, DDX41, del(5q), and bi‐TET2 subgroups experienced more fa- vorable outcomes, although without reaching statistical significance (median OS: 47.9, 44.2, 33.0, and 27.9 months, respectively; P values ranging from 0.17 to 0.96). In contrast, patients in the TP53‐complex and−7/SETBP1 subgroups showed significantly inferior survival, with median OS of 5.7 months (P < 0.001) and 10.1 months (P = 0.001), respectively (Figure 2A). In a multivariable model including age, sex, hemoglobin, neutrophil, and platelet counts, and bone marrow blast percentage as covariates, MDS taxonomy retained an independent and significant as- sociation with post‐HMA OS only in the TP53‐complex (HR 1.78; P < 0.001) and −7/SETBP1 (HR 1.70; P < 0.001) groups. In HMA‐treated patients, baseline clinical risk was heterogeneously distributed across taxonomy groups, with substantial differences in IPSS‐M risk categories and bone marrow blast burden (Supporting Information S1: Appendix Table S2), indicating that treatment allocation and outcomes are influ- enced by clinical factors beyond molecular taxonomy. We then analyzed outcomes in patients who underwent HSCT (1317 out of 5094; 25.9%). The mNOS subgroup had the most favor- able posttransplant outcome, with median OS not reached (P = 0.98). Conversely, the ZRSR2 and TP53‐complex groups showed the poorest survival after HSCT, with median OS of 10.2 months (P = 0.039) and 13.5 months (P < 0.001), respectively (Figure 2B). The prognostic re- levance of MDS taxonomy post‐HSCT was further evaluated in a multivariable model including recipient age and sex, time from diagnosis to transplant, donor type, disease status at transplant (active/pro- gressive disease vs. complete remission), and conditioning regimen (reduced‐intensity vs. standard) as covariates. In this analysis, MDS taxonomy was not independently associated with post‐HSCT OS (P = 0.64). However, relapse rates were significantly higher in the TP53‐complex group (36.1%; P < 0.001) (Figure 2C), and this association re- mained significant in multivariable analysis (HR 2.39; P = 0.001).Transcriptomic profile of genetically‐defined taxonomy subgroupsThe transcriptomic profile of CD34+ bone marrow progenitors was analyzed in a prospective cohort of 447 patients (demographic and clinical characteristics are provided in the Table 1). Consistent with previous studies,18–20 the CD34+ transcriptome reflected the distribution of hematopoietic progenitor cells within the hematopoietic stem and progenitor cell compartment, with genes highly expressed in erythroid and granulocytic myeloid precursors accounting for 33% of total variability. Supervised differential gene expression analysis identified 1187 genes (P < 0.001) that were dif- ferentially expressed across MDS taxonomy subgroups. Unsupervised clustering analysis demonstrated distinct gene expression trajectories among AML‐like, TP53‐complex, and SF3B1 subgroups, which showed a tendency to cluster separately (Figure 3A). To further investigate the relationship between transcriptomic profiles and genotype, a random forest‐based classification algorithm was applied to predict MDS taxonomy from differentially expressed genes. The algorithm′s accuracy varied across subgroups, with recall rates of 94%, 88%, 75%, and 54% for AML‐like, TP53‐complex,SF3B1, and biTET2 groups, respectively (Figure 3B). Classification accuracy was notably lower for other subcategories, likely due to high transcriptomic heterogeneity or, in some cases, small sample sizes (e.g., the EZH2 plus ASXL1 and ZRSR2 groups, each with fewer than HemaSphere | 10 of 17 ### Page 11 (A) (B)F I G U R E 2 Association between MDS Taxonomy molecular groups and clinical outcomes after disease‐modifying therapies, including OS after hypomethylating agents (HMA) (A), OS after allogeneic stem cell transplantation (HSCT) (B), and relapse rate after HSCT (C). Diamonds represent the median OS with horizontal blue bars indicating the 95% confidence intervals. Global differences among molecular groups were assessed using the Log‐Rank test for OS. Hazard ratios (HRs) with 95% CIs reported for each group were derived from Cox proportional hazards models, using the “No‐event” group as reference. MDS, myelodysplastic syndromes/neoplasms; OS, overall survival.11 of 17 | ### Page 12 10 patients, and DDX41 that is not represented in this cohort). Overall, these findings suggest that a clear distinct gene signature characterizes AML‐like, TP53‐complex, SF3B1, and biTET2 subgroup, while higher transcriptome heterogeneity was observed across other genetic entities (Figure 3C). Noteworthy, pathway analysis found restricted gene‐set enrich- ment in these groups. Heme and porphyrins metabolism pathways were enriched for SF3B1 subgroup (Figure 3D). The TP53‐complex subgroup exhibited dysregulation of genes associated with interferon signaling, alongside pathways involved in platelet aggregation and activation. In contrast, innate immunity pathways—such as Toll‐like receptor signaling—were more specifically enriched in the bi‐TET2subgroup. A markedly more heterogeneous set of enriched pathways was observed in the AML‐like subgroup, which included reactive oxygen species and inflammasome‐related pathways, uniquely asso- ciated with this group. Additionally, innate immunity and erythroid precursor pathways were identified in AML‐like subgroups over- lapping with the bi‐TET2 and SF3B1 categories, respectively (Figure 3D). Overall, the AML‐like subgroup displayed a more in- tricate transcriptomic landscape, consistent with its classification as the final evolutionary stage of several lower‐risk, molecularly defined myeloid neoplasms.DISCUSSIONThe study of hematological neoplasms has been transformed by genomic annotation. While morphology remains a fundamental tool for diagnosing hematologic neoplasms and serves as a robust basis for future classification, it is now well recognized that dysplastic mor- phology results from dysregulation of blood cell production, differ- entiation, or survival that is itself driven by gene mutations or altered gene expression.1,2,6,7,26 Somatic mutations and cytogenetic aberra- tions have already been incorporated into a robust risk‐stratification system for MDS, the IPSS‐M.14 However, risk‐stratification predicts patient outcome in an incremental manner, without defining unique biologic disease entities.14,26 The integration of molecular information into classification systems for hematological neoplasms has the po- tential to define distinct genomic subgroups with different ther- apeutic implications, responses, and outcomes; to identify biomarkers(C)F I G U R E 2 (Continued) HemaSphere | 12 of 17 ### Page 13 (A) (B) (C) (D)F I G U R E 3 Transcriptomic profile of genetically‐defined Taxonomy subgroups (prospective cohort, n = 447). (A) Similarity matrix displaying pairwise Pearson correlation coefficients for all samples. Taxonomy and WHO classification categories are presented as covariates in the top annotation. Hierarchical clustering reveals distinct transcriptomic divergence among the SF3B1, TP53‐complex, and AML‐like subgroups, predominantly enriched in the left, middle, and right sections of the dendrogram, respectively. (B) Results of a random forest classifier trained on gene expression data to predict taxonomy subgroups. Recall metrics stratified by taxonomy group indicate higher specificity in transcriptomic profiles for AML‐like, TP53‐complex, bi‐TET2, and SF3B1 subgroups compared to other classifications.(C) Heatmap illustrating the expression levels of differentially expressed genes across taxonomy subgroups (AML‐like, bi‐TET2, SF3B1, TP53‐complex, and remaining subgroups). Rows represent individual genes (n = 1187), while columns denote samples (n = 447). Covariates indicate taxonomy classifications and WHO subtypes.(D) Results of pathway enrichment analysis for differentially expressed genes in the AML‐like, bi‐TET2, SF3B1, and TP53‐complex subgroups. The x‐axis represents the gene count per pathway, while the y‐axis lists enriched pathways (q‐value < 0.05). The bar color corresponds to the q‐value, indicating statistical significance. AML, acute myeloid leukemia.13 of 17 | ### Page 14 for disease monitoring; and to ultimately lay the foundation for per- sonalized treatments.7 Since genomic profiling has become progres- sively implemented into the diagnostic work‐up of hematologic neoplasms, disease classification is now evolving into an essential tool to drive personalized medicine.7,12,27 Recently, the IWG‐PM established a refined molecular classifi- cation system, termed MDS taxonomy, to better reflect the under- lying biology of these neoplasms.13 MDS taxonomy validated the clinical relevance of 5 well‐established entities (MDS with SF3B1,TP53, del(5q), and AML‐like subtypes), confirmed the clinical sig- nificance of 3 previously reported groups (biTET2, CCUS‐like, and MDS with der(1;7)) and introduced 8 novel subgroups with distinct molecular features (−7/SETBP1, EZH2‐ASXL1, IDH‐STAG2, BCOR/L1, U2AF1, SRSF2, and ZRSR2).13 In this study, we provided a large, in- dependent confirmation that MDS taxonomy defines subgroups with distinct clinical features and outcomes, both in terms of natural his- tory and within specific therapeutic subgroups. Although the purpose of MDS taxonomy differs from that of prognostic models such as IPSS‐M,14 both aim to inform clinical decision‐making.13,14,26 We confirmed that IPSS‐M outperforms MDS taxonomy in predicting OS and AML transformation risk. Additionally, we assessed whether incorporating gene interactions as defined by MDS taxonomy could refine the accuracy of IPSS‐M. Our findings indicate that gene interactions featured in the MDS taxonomy do not significantly enhance the predictive accuracy of the conventional models. Instead, IPSS‐M effectively stratified patients within MDS taxonomy subgroups to different survival probabilities. We next explored whether the MDS taxonomy could provide additional insights into the management of patients receiving disease‐modifying therapies, with the caveat that the retrospective nature of the available data may introduce potential bias due to variability in therapeutic decision‐making across centers and over time. HMA is the only class of drugs approved for the treatment of higher‐risk MDS in patients ineligible for transplantation.4,5 However, only 40%–50% of patients experience hematologic improvement, and complete response rate is 10‐15%.28 Predictive models based on clinical features alone are not sufficiently conclusive to withhold HMA treatment from eligible patients,29,30 and IPSS‐M has recently failed to stratify individual response probability—although OS was inversely related to IPSS‐M risk.26 We observed that MDS taxonomy, despite incorporating genomic information in a different way with respect to IPSS‐M, did not predict individual response probability to HMA. These findings suggest that while genomic information captured by IPSS‐M and MDS taxonomy reflects the disease biology and the ag- gressiveness of MDS subtypes, additional factors beyond gene mu- tations can influence HMA sensitivity.31,32 Allogeneic HSCT is the only potentially curative option for MDS patients, yet disease relapse remains the leading cause of failure.4,5 Genomic data incorporated into IPSS‐M has been shown to predict relapse risk, thereby refining patient management.26,33 MDS tax- onomy did not significantly improve the identification of patients at high risk of transplant failure, except for the TP53‐complex group, which has been known to be associated with high relapse risk fol- lowing HSCT.34,35 The MDS taxonomy framework aims to provide a foundation for a more biologically refined classification of these diseases. In addition, it could serve as a stepping stone for future research to generate new hypotheses for the development of subtype‐specific treatments and may help refine inclusion criteria in next‐generation clinical trials, thereby identifying more homogeneous patient populations com- pared with conventional clinical stratification.13 In this context, our study incorporated RNA sequencing as an orthogonal technology to further validate the biological distinctiveness of MDS taxonomy groups and the transcriptional consistency within each group. Although some molecular sub- groups (e.g., DDX41, EZH2‐ASXL1, and ZRSR2) were under- represented, transcriptomic profiling of CD34+ cells provided preliminary evidence supporting the biological relevance of the proposed taxonomy. To investigate the relationship between transcriptome and gen- otype, we applied a random forest‐based classification algorithm using differentially expressed genes to predict MDS taxonomy cate- gories. The AML‐like, bi‐TET2, SF3B1, and TP53‐complex groups showed recall rates >70%, while classification performance was lower in other subgroups, likely due to greater transcriptomic heterogeneity. Gene ontology analyses revealed significant pathway enrichment in specific MDS taxonomy categories: the SF3B1 subgroup demon- strated robust activation of genes involved in erythroid differentia- tion, consistent with previous findings.18–20 The TP53‐complex group showed dysregulation of genes associated with interferon signaling, highlighting the central role of TP53 gene in modulating immune re- sponses.36 Recent multi‐omics single‐cell analyses of bone marrow cells in TP53‐mutated MDS further confirmed a specific immune dysregulation in the tumor microenvironment, characterized by myeloid‐driven inflammation and impaired antigen presentation, which may drive the dismal prognosis found in these patients.17 Pathways related to innate immunity—such as Toll‐like receptor signaling—were specifically enriched in the bi‐TET2 group, supporting the emerging role of this gene in regulating immune homeostasis and contributing to clonal evolution in MDS.37 Finally, the AML‐like subgroup displayed a markedly heterogeneous pattern of enriched pathways, including those related to reactive oxygen species and in- flammasome activation—features uniquely associated with this group.38 Notably, innate immunity and erythroid precursor signatures in AML‐like groups overlapped with the bi‐TET2 and SF3B1 cate- gories, respectively, indicating a more complex and composite tran- scriptomic landscape that reflects the transitional evolutionary status of the AML‐like group from molecularly defined subtypes of low‐risk MDS.39,40 Overall, these findings suggest that transcriptome heterogeneity varied across the genetic groups, with distinct gene signature evident for AML‐like, biTET2, SF3B1, and TP53‐complex subgroups. Further studies to validate these findings and refine multi‐omics integration approaches are warranted to establish a mechanistic classification of MDS that could be an innovative reference model to improve clinical decision making, optimize the design of clinical trials and potentially guide the development of innovative treatments.7A U THO R CO NTR I BU TI ON S Giulia Maggioni: Conceptualization; investigation; writing—original draft; formal analysis; data curation. Gabriele Todisco: Con- ceptualization; investigation; writing—original draft; formal analysis; data curation. Elisabetta Sauta: Conceptualization; investigation; writing—original draft; formal analysis; data curation. Luca Lanino: Conceptualization; investigation; formal analysis; data curation.Somedeb Ball: Writing—review and editing; data curation. Jan P. Bewersdorf: Writing—review and editing; data curation. Tariq Kewan: Writing—review and editing; data curation. Najla H. Al Ali: Writing—review and editing; data curation. Pierre Fenaux: Writing—review and editing; data curation. Uwe Platzbecker: Writing—review and editing; data curation. Valeria Santini: Writing—review and editing; data curation. Zhuoer Xie: Writing—review and editing; data curation.Andrew M. Brunner: Writing—review and editing; data curation.Andrew T. Kuykendall: Writing—review and editing; data curation.John M. Bennett: Writing—review and editing; data curation. HemaSphere | 14 of 17 ### Page 15 Rena Buckstein: Writing—review and editing; data curation. Rafael Bejar: Writing—review and editing; data curation. Hetty E. Carraway: Writing—review and editing; data curation. Amy E. DeZern: Writing—review and editing; data curation. Elizabeth A. Griffiths: Writing—review and editing; data curation. Stephanie Halene: Writing—review and editing; data curation. Robert P. Hasserjian: Writing—review and editing; data curation. Alan F. List: Writing—review and editing; data curation. Sanam Loghavi: Writing—review and editing; data curation.Olatoyosi Odenike: Writing—review and editing; data curation. Eric Padron: Writing—review and editing; data curation. Mrinal M. Patnaik: Writing—review and editing; data curation. Gail J. Roboz: Writing—review and editing; data curation. Maximilian Stahl: Writing—review and editing; data curation. Mikkael A. Sekeres: Writing—review and editing; data curation. David P. Steensma: Writing—review and editing; data curation. Michael R. Savona: Writing—review and editing; data curation. Justin Taylor: Writing—review and editing; data curation. Mina L. Xu: Writing—review and editing; data curation. David A. Sallman: Writing—review and editing; data curation. Stephen D. Nimer: Writing—review and editing; data curation. Christopher S. Hourigan: Writing—review and editing; data curation. Andrew H. Wei: Writing—review and editing; data curation.Alessia Campagna: Writing—review and editing; data curation. Marta Ubezio: Writing—review and editing; data curation. Elena Riva: Writing—review and editing; data curation. Denise Ventura: Writing—review and editing; data curation. Nicole Pinocchio: Writing—review and editing; data curation. Matteo Zampini: Writing—review and editing; data curation. Alessandro Buizza: Writing—review and edit- ing; data curation. Antonio Russo: Writing—review and editing; data curation. Francesco Pesce: Writing—review and editing; data cura- tion. Saverio D'Amico: Writing—review and editing; data curation.Gianluca Asti: Writing—review and editing; data curation. Mattia Delleani: Writing—review and editing; data curation. Erica Travaglino: Writing—review and editing; data curation. Cristina A. Tentori: Writing—review and editing; data curation. Ivan Ferrari: Writing—review and editing; data curation. Alessandra Crespi: Writing—review and editing; data curation. Giulia Figini: Writing—review and editing; data curation. Maria Di Matteo: Writing—review and editing; data curation. Matteo Brindisi: Writing—review and editing; data curation.Nicla Manes: Writing—review and editing; data curation. Chiara Milanesi: Writing—review and editing; data curation. Laura Crisafulli: Writing—review and editing; data curation. Francesca Ficara: Writing—review and editing; data curation. Gastone Castellani: Writing—review and editing; data curation. Uma M. Borate: Writing—review and editing; data curation. Fabio Efficace: Writing—review and editing; data curation. Steven D. Gore: Writing—review and editing; data curation. Tae K. Kim: Writing—review and editing; data curation.Maria Diez‐Campelo: Writing—review and editing; data curation.Arjan A. van de Loosdrecht: Writing—review and editing; data cura- tion. Navel Daver: Writing—review and editing; data curation. Maria Rozman: Writing—review and editing; data curation. Alberto Orfao: Writing—review and editing; data curation. Sa A. Wang: Writing—review and editing; data curation. Kathryn Foucar: Writing—review and editing; data curation. Marcelo Iastrebner: Writing—review and editing; data curation. Phillip Scheinberg: Writing—review and edit- ing; data curation. Yasushi Miyazaki: Writing—review and editing; data curation. Yazan F. Madanat: Writing—review and editing; data curation. Aref Al‐Kali: Writing—review and editing; data curation.Moshe Mittelmann: Writing—review and editing; data curation.Thomas Cluzeau: Writing—review and editing; data curation. Lionel Ades: Writing—review and editing; data curation. Ulrich Germing: Writing—review and editing; data curation. Guillermo Garcia‐Manero: Writing—review and editing; data curation. Shahram Kordasti: Writing—review and editing; data curation. Torsten Haferlach: Writing—review and editing; data curation. Amer M. Zeidan: Writing—review and editing; supervision; writing—original draft; conceptualization; funding acquisition; visualization. Rami S. Kom- rokji: Writing—review and editing; supervision; writing—original draft; conceptualization; funding acquisition; visualization. Matteo G. Della Porta: Writing—review and editing; supervision; writing—original draft; conceptualization; funding acquisition; visualization.CONFLICT OF INTERES T S TATEMENTU.G. reports speaker honoraria from Novartis, AbbVie, and BMS; and institutional research support from BMS, AbbVie, and Jazz Pharma- ceuticals. F.E. reports consultancy or advisory roles for AbbVie, Incyte, Syros, Novartis, and Jazz Pharmaceuticals. S.H. reports research sup- port from STORM Therapeutics and AstraZeneca. E.A.G. reports hon- oraria from AAMDS, MedscapeLIVE!, MediCom Worldwide, MJH Life Sciences, ASH, MDS International Foundation, and Physicians' Educa- tion Resource; consulting fees from AbbVie, Alexion, Apellis, Takeda Oncology, Astex/Taiho Oncology, Alexion/AstraZeneca Rare Disease, Celgene/BMS, CTI BioPharma, Novartis, Partner Therapeutics, Picnic Health, and Servier; and research funding from Alexion, Apellis, Astex/ Otsuka/Taiho Oncology, Blueprint Medicines, Celldex Therapeutics, Genentech, and NextCure. M.A.S. reports participation on advisory boards for BMS, Kurome, Schrödinger, and Karyopharm. M.D‐C. re- ports participation on a data safety monitoring board or advisory board for BMS, Novartis, Blueprint Medicines, GSK, Agios, Hemavan, Syros, Keros, Curis, and Astex/Otsuka; and payment or honoraria for lectures, presentations, speakers' bureaus, manuscript writing, or educational events for BMS, Novartis, and Keros. U.P. reports research support and honoraria from BMS, Geron, Curis, AbbVie, and Janssen. R.Be. reports employment or equity from Aptose Biosciences; participation on ad- visory boards for BMS, Servier, NeoGenomics, and Geron; being Data Monitoring Committee Chair for Gilead, Ipsen, and Keros; and con- sultancy for TenSixteen. YM reports honoraria from Nippon Shinyaku, BMS, Novartis, Sumitomo Pharma, Kyowa Kirin, AbbVie, Daiichi San- kyo, Takeda, Janssen, Astellas, Pfizer, Eisai, and Otsuka; and research funding from Chugai. A.E.D. reports participation on advisory boards, consultancy, or honoraria from Celgene/BMS, Agios, Novartis, Astellas, and Gilead; and participation on clinical trial committees or data safety monitoring boards for Novartis, AbbVie, Kura, Geron, Servier, Keros, and Celgene/BMS. D.P.S. reports employment by Ajax Therapeutics; former employment by Novartis; and minor equity in Arrowhead and Bluebird. T.K.K. reports consultancy for Agenus and ImmunoBiome. A.A. reports research supprot to institution from Novartis, Celgen/ BMS, ALX Oncology, Astex/Taiho Oncology, AstraZeneca/Med- Immune, H3BioMedicine/Hemavant, Aprea; Steering committee membership support from Novartis. All other authors declare no competing interests.D A TA AV A IL AB I L IT Y S TA TE M E N TThe data that support the findings of this study are available on re- quest from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.FUNDINGThe study was supported by: European Union—Horizon 2020/2023 program and Innovative Health Initiative, IHI (GenoMed4All project #101017549 to M.G.D.P., U.P., M.D‐C., P.F., T.H., G.C.; Synthema pro- ject, #101095530 to M.G.D.P.; Synthia project #101172872 to M.G.D.P., S.DA., G.C; RealiseD project #101165912 to M.G.D.P); European Union—Next Generation EU—NRRP M6C2 (Investment 2.1 Enhance- ment and strengthening of biomedical research in the NHS—Project15 of 17 | ### Page 16 PNRR‐MAD‐2022‐12376695 and PNRR‐MCNT1‐2023‐12377648 to M.G.D.P.); AIRC Foundation (Associazione Italiana per la Ricerca contro il Cancro, Milan, Italy – Project #29483 to M.G.D.P.; #26537 to V.S.; #28131 to G.M., #31589 to E.R., #31032 to G.T.); 5 × 1000 MYNERVA (Myeloid Neoplasms Research Venture AIRC—project #21267 to M.G.D.P); PRIN (Ministry of University & Research, Italy—Project 20229B28PE to M.G.D.P); TRAIN srl, Milan, Italy (to S.DA; M.D.). Spanish Ministry of Health, Instituto de Salud Carlos III, UmbrellaSumma project PI23/01103 and Umbrella project, PI20/00970 (to M.D.‐C).SU PPOR TI NG I NF OR MAT I ONAdditional supporting information can be found in the online version of this article.ORCIDAlessandro Buizza https://orcid.org/0009-0008-5764-1964Francesca Ficara https://orcid.org/0000-0003-1873-1189Fabio Efficace https://orcid.org/0000-0002-5065-5166Maria Diez‐Campelo https://orcid.org/0000-0002-1467-6779Alberto Orfao https://orcid.org/0000-0002-0007-7230Sa A. Wang https://orcid.org/0000-0001-9385-9911Thomas Cluzeau https://orcid.org/0000-0002-6745-1127Rami S. Komrokji https://orcid.org/0000-0002-1876-5269Matteo G. Della Porta https://orcid.org/0000-0002-6915-5970REF EREN CES1. Cazzola M. Myelodysplastic syndromes. N Engl J Med. 2020;383(14): 1358‐1374. 2. Cazzola M, Della Porta MG, Malcovati L. The genetic basis of mye- lodysplasia and its clinical relevance. Blood. 2013;122(25):4021‐4034. 3. Arber DA, Orazi A, Hasserjian R, et al. The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood. 2016;127(20):2391‐2405. 4. Malcovati L, Hellström‐Lindberg E, Bowen D, et al. Diagnosis and treatment of primary myelodysplastic syndromes in adults: re- commendations from the European LeukemiaNet. Blood. 2013; 122(17):2943‐2964. 5. Fenaux P, Haase D, Santini V, Sanz GF, Platzbecker U, Mey U. Myelodysplastic syndromes: ESMO clinical practice guidelines for diagnosis, treatment and follow‐up. Ann Oncol. 2021;32(2):142‐156. 6. Bersanelli M, Travaglino E, Meggendorfer M, et al. Classification and personalized prognostic assessment on the basis of clinical and genomic features in myelodysplastic syndromes. J Clin Oncol. 2021;39(11):1223‐1233. 7. Cazzola M, Sehn LH. Developing a classification of hematologic neoplasms in the era of precision medicine. Blood. 2022;140(11): 1193‐1199. 8. Konorkji R, Lanino L, et al. Data‐driven harmonized classification system for myelodysplastic syndromes; a consensus paper from the International Consortium for Myelodysplastic Syndromes (icMDS).Lancet Haematol. 2024;11(11):e862‐e872. 9. Khoury JD, Solary E, Abla O, et al. The of the World Health Organization Classification of Haematolymphoid Tumours: Myeloid and Histiocytic/ dendritic Neoplasms. 5th ed Leukemia. 2022;36(7):1703‐1719. 10. Arber DA, Orazi A, Hasserjian RP, et al. International Consensus Classification of Myeloid Neoplasms and Acute Leukemias: in- tegrating morphologic, clinical, and genomic data. Blood. 2022; 140(11):1200‐1228. 11. Bernard E, Nannya Y, Hasserjian RP, et al. Implications of TP53 allelic state for genome stability, clinical presentation and outcomes in myelodysplastic syndromes. Nat Med. 2020;26(10):1549‐1556. 12. Della Porta MG, Bewersdorf JP, Wang YH, Hasserjian RP. Future directions in myelodysplastic syndromes/neoplasms and acute myeloid leukaemia classification: from blast counts to biology.Histopathology. 2025;86(1):158‐170. 13. Bernard E, Hasserjian RP, Greenberg PL, et al. Molecular taxonomy of myelodysplastic syndromes and its clinical implications. Blood. 2024;144(15):1617‐1632. 14. Bernard E, Tuechler H, Greenberg PL, et al. Molecular international prognostic scoring system for myelodysplastic syndromes. NEJM Evid. 2022;1(7):EVIDoa2200008. 15. Greenberg PL, Tuechler H, Schanz J, et al. Revised international prognostic scoring system for myelodysplastic syndromes. Blood. 2012;120(12):2454‐2465. 16. Papaemmanuil E, Gerstung M, Malcovati L, et al. Clinical and biolo- gical implications of driver mutations in myelodysplastic syndromes.Blood. 2013;122(22):3616‐3627. 17. Zampini M, Riva E, Lanino L, et al. Characterization and Clinical Im- plications of p53 Dysfunction in Patients with Myelodysplastic Syndromes. J Clin Oncol. 2025;43(18):2069‐2083. 18. Shiozawa Y, Malcovati L, Gallì A, et al. Gene expression and risk of leukemic transformation in myelodysplasia. Blood. 2017;130(24): 2642‐2653. 19. Pellagatti A, Armstrong RN, Steeples V, et al. Impact of spliceosome mutations on RNA splicing in myelodysplasia: dysregulated genes/ pathways and clinical associations. Blood. 2018;132(12):1225‐1240. 20. Todisco G, Creignou M, Bernard E, et al. Integrated genomic and transcriptomic analysis improves disease classification and risk stra- tification of MDS with ring sideroblasts. Clin Cancer Res. 2023;29(20): 4256‐4267. 21. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA‐seq data with DESeq2. Genome Biol. 2014;15(12):550. 22. Kuhn M. Building predictive models in R using the caret package.J Stat Softw. 2008;28(5):1‐26. 23. Wu T, Hu E, Xu S, et al. clusterProfiler 4.0: a universal enrichment tool for interpreting omics data. Innovation (Cambridge (Mass.)). 2021;2(3):100141. 24. Cheson BD. Clinical application and proposal for modification of the International Working Group (IWG) response criteria in myelodys- plasia. Blood. 2006;108(2):419‐425. 25. Harrell FE. Evaluating the yield of medical tests. JAMA. 1982; 247(18):2543‐2546. 26. Sauta E, Robin M, Bersanelli M, et al. Real‐world validation of mo- lecular international prognostic scoring system for myelodysplastic syndromes. J Clin Oncol. 2023;41(15):2827‐2842. 27. Duncavage EJ, Bagg A, Hasserjian RP, et al. Genomic profiling for clinical decision making in myeloid neoplasms and acute leukemia.Blood. 2022;140(21):2228‐2247. 28. Fenaux P, Mufti GJ, Hellstrom‐Lindberg E, et al. Efficacy of azaciti- dine compared with that of conventional care regimens in the treatment of higher‐risk myelodysplastic syndromes: a randomised, open‐label, phase III study. Lancet Oncol. 2009;10(3):223‐232. 29. Zeidan AM, Sekeres MA, Garcia‐Manero G, et al. Comparison of risk stratification tools in predicting outcomes of patients with higher‐risk myelodysplastic syndromes treated with azanucleo- sides. Leukemia. 2016;30(3):649‐657. 30. Kuendgen A, Müller‐Thomas C, Lauseker M, et al. Efficacy of aza- citidine is independent of molecular and clinical characteristics—an analysis of 128 patients with myelodysplastic syndromes or acute myeloid leukemia and a review of the literature. Oncotarget. 2018;9(45):27882‐27894. 31. Follo MY, Finelli C, Mongiorgi S, et al. Reduction of phosphoinositide‐phospholipase C beta1 methylation predicts the responsiveness to azacitidine in high‐risk MDS. Proc Natl Acad Sci USA. 2009;106(39):16811‐16816. HemaSphere | 16 of 17 ### Page 17 32. Liu YC, Kwon J, Fabiani E, et al. Demethylation and up‐regulation of an oncogene after hypomethylating therapy. N Engl J Med. 2022;386(21):1998‐2010. 33. Gurnari C, Robin M, Adès L, Aljurf M, Almeida AM, et al. Clinical‐genomic profiling of MDS to inform allo‐HSCT: recommendations from an international panel on behalf of the EBMT. Blood. 2025;145(18):1087‐2001. 34. Della Porta MG, Gallì A, Bacigalupo A, et al. Clinical effects of driver somatic mutations on the outcomes of patients with myelodysplastic syndromes treated with allogeneic hematopoietic stem‐cell trans- plantation. J Clin Oncol. 2016;34(30):3627‐3637. 35. Tentori CA, Gregorio C, Robin M, et al. Clinical and genomic‐based decision support system to define the optimal timing of allogeneic hematopoietic stem‐cell transplantation in patients with myelodysplastic syndromes. J Clin Oncol. 2024;42(24): 2873‐2886. 36. Sallman DA, McLemore AF, Aldrich AL, et al. TP53 mutations in myelodysplastic syndromes and secondary AML confer an im- munosuppressive phenotype. Blood. 2020;136(12):2812‐2823. 37. Jiang S. Tet2 at the interface between cancer and immunity. Commun Biol. 2020;3(1):667. 38. Corradi G, Bassani B, Simonetti G, et al. Release of IFNγ by acute myeloid leukemia cells remodels bone marrow immune micro- environment by inducing regulatory T Cells. Clin Cancer Res. 2022;28: 3141‐3155. 39. Cheng WY, Li JF, Zhu YM, et al. Transcriptome‐based molecular subtypes and differentiation hierarchies improve the classification framework of acute myeloid leukemia. Proc Natl Acad Sci USA. 2022;119(49):e2211429119. 40. Docking TR, Parker JDK, Jädersten M, et al. A clinical transcriptome approach to patient stratification and therapy selection in acute myeloid leukemia. Nat Commun. 2021;12(1):2474.17 of 17 |