Integrated multi-omic profiling identifies molecular subtypes and predicts outcomes in newly diagnosed peripheral T-cell lymphomas

Introduction Peripheral T-cell lymphomas (PTCLs) comprise a heterogeneous group of aggressive hematological malignancies, accounting for 10–15% of all Non-Hodgkin Lymphomas (NHLs). Despite therapeutic advances, prognosis re mains poor for most patients. In this context, the PTCL-13 study was designed to evaluate the addition of Romidepsin to anthracycline-based chemotherapy (Ro-CHOEP) followed by autologous stem cell transplantation (SCT) (NCT02223208). No tably, the clinical analysis of this cohort found no signicant improvement by adding romidepsin (Chiappella et al. Leukemia, 2023), highlighting the need for new drugs as well as molecular proling to guide patient stratication and possibly thera peutic decision-making. To address this, we performed an integrative multi-omics analysis of patients enrolled in the PTCL13 clinical trial with the aim of identifying molecular signatures associated with outcomes. Material and MethodsFormalin-xed, parafn-embedded (FFPE)tissueandplasmawereprospectivelycollectedatbaseline from 73 newly diagnosed PTCL patients with different histologies [nodal T-follicular helper lymphoma (nTFHL;n=27); Periph eral T-cell lymphoma not otherwise specied (PTCL-NOS;n=29); Anaplastic Lymphoma Kinase-negative Anaplastic large-cell 3NOVEMBER2025|VOLUME146,NUMBERSupplement1 ©2025AmericanSocietyofHematology.PublishedbyElsevier Inc. All rights reserved. Downloaded from ashpublications.​org/​blood/​article-pdf/​146/​Supplement 1/​1758/​2447171/​blood-4505-main.​pdf by guest on 03 September 2026 1758 POSTER Session 621. Lymphomas: Translational– Molecular and Genetic lymphoma(ALKneg ALCL;n=17)]. Targeted sequencing was performed on paired FFPE andplasmasamples(Illumina NextSeq 550) using a panel of 60 genes (Roche) to identify somatic mutations. RNA from FFPE was used to perform gene expres sion proling using the nCounter 780-gene panel (NanoString) and bulk RNA-sequencing (Illumina NextSeq 550 platform). Clustering analysis of multi-omic datasets was performed on R (version 4.5.0). Results Unsupervised clustering of transcriptomic data identied three distinct clusters (Cluster A, Cluster B, and Cluster C). Cluster A had signicantly higher proportion of nTFHL patients (58%; p<0.0001), but also included some PTCL-NOS (35%) and ALKneg ALCL (7%) patients. Consistent with this, genes associated with T-follicular helper (TFH) and B-cells (TCF7, CD22, MS4A1) (adjusted p<0.01) were signicantly upregulated in Cluster A. Interestingly, there was also a signicant downregula tion of genes associated with myeloid cells, including CD14, CD163, and CD33 (adjusted p<0.05) in Cluster A. Conversely, genes related to M2 macrophages (CCL13, CCL17) (adjusted p<0.05) and exhausted T cells (PDL2, CD276) (adjusted p<0.05) were signicantly upregulated in Clusters B and C, respectively. Notably, patients in Cluster A had superior progression-free survival (PFS) as compared to patients in Clusters B and C (p<0.05). To better dene the molecular landscape of PTCLs, an in tegrated multi-omic analysis combining genomic and transcriptomic proles was conducted. The Cluster-of-Clusters analysis (COCA) approach identied two principal clusters (Cluster 1 and Cluster 2). Cluster 1 closely mirrored the transcriptional and histological prole of Cluster A. Furthermore, there was a higher frequency of gene mutations in epigenetic regulators, in cluding TET2 (36%), RHOA (18%), IKZF2 (14%), and DNMT3A (14%), and a signicant co-occurrence of RHOA and TET2 in this group (p<0.01). In contrast, Cluster 2 closely resembled Clusters B and C in terms of gene expression prole and histological subgrouping. Additionally, mutations in oncogenic drivers such as CARD11 (7%), TP63 (4%) and MYC (4%) with signicant co occurrence between CARD11 and NOTCH1 genes (p<0.01) were found to be associated with Cluster 2. Patients in Cluster 1 had signicantly higher complete response (CR) rates to Ro-CHOEP as comparedtoCluster 2 (46%CRinCluster 1vs. 18%CR in Cluster 2; p<0.05). Additionally, Cluster 1 patients demonstrated a trend toward prolonged PFS (p=0.06), and signicantly improved overall survival (p<0.05). Conclusions Multi-omic clustering identied a subset of patients likely to benet from Ro-CHOEP treatment,characterizedby a less immunosuppressive tumor microenvironment and higher frequency of mutation in epigenetic regulators. Collectively, our ndings underscore the potential of integrative molecular proling as a tool to guide risk stratication and provide bio logical insights beyond conventional histological sub-classication. Ongoing studies integrating bulk RNA sequencing and whole genome sequencing data into this model will be presented, allowing us to further rene molecular characterization of PTCLs.