Pre-vaccine immune profiling in cancer patients identifies correlates of COVID-19 vaccine responses Authors: Maryam Kazerani, Simeon Mahov, Joslyn Foley, Noah M Merin, So Yung Choi, Joseph Lownik, Alexander M Xu, Larry Milshteyn, Anton Luis Villamejor, Maimoona Nadri, Nathalie Nguyen, Elham Kazemian, Omid Hamid, Justin Darrah, Jun Gong, Ronald Paquette, Robert Vescio, Inderjit Mehmi, Warren G Tourtellotte, V Krishnan Ramanujan, Karen L Reckamp, Jane C Figueiredo, Akil Merchant Published: 2025 ABSTRACT ABSTRACTmRNA vaccines have been highly effective against SARS-CoV-2-related severe illness and are currently undergoing investigation as anti-cancer therapeutics. Our prior work has shown that patients undergoing cancer therapy have reduced immune responses to mRNA-based vaccines. Therefore, further investigation into immunologic responses in the setting of immune-altered hosts is warranted. In this study, we investigated pre-vaccination peripheral immune cell repertoire from a cohort of 66 patients with cancer on active treatment. Immune cell repertoire was characterized by mass cytometry to identify key immune subsets for COVID-19 vaccine response. Immunological populations were then assessed for their association with spike-specific antibody titers measured by ELISA and the breadth and depth of T-cell response to vaccine by TCR sequencing. Immune features significantly correlated with response were selected using an iterative bootstrapping model. We identified immunological features including abundance and functional state of T and B cells, expression of co-stimulatory and -inhibitory molecules, and presence of innate lymphoid cells, and myeloid-derived suppressor cells correlated with humoral and cellular responses. Our findings suggest that vaccine-induced immune responses could serve as biomarkers of immune fitness, in general and specifically for patients receiving anti-cancer mRNA vaccines, providing insights to tailor future therapeutic strategies for immune-altered patients.KEYWORDS: Immunophenotyping, immune fitness, cancer, humoral response, TCR response, vaccine, SARS-CoV-2 Introduction Patients receiving anti-cancer treatments experience immunosuppression due to the cancer disease itself and the effects of anti-cancer therapies, making them more susceptible to SARS-CoV-2 infection and at higher risk for severe outcomes.1,2 Immunization with COVID-19 vaccines and booster uptakes confers protection by inducing the humoral and cellular responses.3 However, previous studies suggest that cancer patients have compromised immune responses to COVID-19 vaccines. More prominently, patients receiving treatments demonstrate reduced humoral and cellular responses, compared to not-on-treatment patients.4,5 It has been shown that treatment with B-cell-directed antibodies and small-molecule inhibitors correlate with impaired humoral responses in patients with hematologic malignancies.6 Reduced T cell response has also been observed in these patients, as well as in those receiving steroid treatment.5,7 In patients with solid tumors, humoral responses are generally comparable to healthy individuals, but the mounted cellular responses are heterogeneous in different patient cohorts.6,8–10 Immune protection conferred by COVID-19 vaccines is mediated by a complex coordinated response of immune components including B-cell, CD4+ T follicular helper (Tfh) cells, plasma cells, memory B cells, and T-cell subsets.11–14 Apart from adaptive immune responses, COVID-19 vaccination provides additional protection by stimulating monocytes and natural killer (NK) cells from the innate immune compartment.15,16 However, in patients with cancer, dysregulation of the immune system may hinder mounting of a protective immune response. Several immune populations with regulatory function like regulatory T (T-reg) and B (B-reg) cells and with immunosuppressive potential like myeloid-derived suppressor cells (MDSC) could potentially attenuate the function of T and B cells, thereby contributing to the systemic immunosuppression in cancer patients.17–19 While previous studies have focused predominantly on immunologic responses in cancer patients during SARS-CoV-2 infection or after COVID-19 vaccinations, there are no clinically useful predictive biomarkers to determine if immunosuppressed cancer patients will respond to vaccines. Here, we assessed the pre-vaccine immune profile of patients with hematologic malignancies, severe aplastic anemia, and solid tumors undergoing treatment. By correlating the immunologic phenotypes with SARS-CoV-2-specific antibody and cellular responses, we identified specific immune populations associated with immune responses after vaccination. Materials and methods Study population Patients with hematologic malignancies, severe aplastic anemia, and solid tumors were enrolled in a prospective cohort study – the U.S. National Cancer Institute-funded Serological Sciences Network (SeroNet)-Coronavirus Risk Associations and Longitudinal Evaluation (CORALE) study – between December 22, 2020, and January 31, 2022. We oversampled patients with altered immunity, including B-cell or plasma cell malignancies, bone marrow transplantation (BMT), and cancer patients receiving immune checkpoint inhibitors (ICI).20–22 Patients who underwent BMT between vaccination and spike-specific antibody titer measurement were excluded. For this study, we included a subset of participants (n = 66) who had no prior history of SARS-CoV-2 infection and provided a blood sample before vaccination. Patients’ demographics and clinical information are summarized in Supplementary Table S1. The study was conducted following the ethical principles outlined in the Declaration of Helsinki. Ethical approval was obtained from Cedars-Sinai Medical Center Institutional Review Board (Approval No. Study00001316). All participants provided written informed consent prior to enrollment. Cytometry by time-of-flight (CyTOF) analysis Peripheral blood mononuclear cells (PBMCs) were collected from patients. The CyTOF phenotyping panel was designed to classify major immune cell types and key subsets involved in the vaccine response. Data were acquired using a Helios™ mass cytometer (Fluidigm, California, U.S.). CyTOF data were batch-corrected using CyCombine,23 followed by a two-stage stacked hierarchical-Leiden clustering approach.24 Lineage-based primary clustering (e.g., CD3, CD19, CD33) defined major compartments (T, B, and non-B/T cells), which were then subclustered with additional phenotypic markers to delineate functionally distinct immune subsets. To construct the full feature set for correlation analysis, the immune populations per patient sample were quantified in relative frequencies and absolute concentrations, calibrated using CBC data. To obtain a more stable estimate of lymphocyte and monocyte frequencies and counts, the mean of serial CBC values collected from 60 d before the CyTOF draw up to 3 d post – first vaccination was included in the analysis. SARS-CoV-2 immunoglobulin G antibody testing Measurement of the antibodies against the receptor-binding domain (RBD) of the S1 subunit of the viral spike protein [IgG(S-RBD)] and viral nucleocapsid protein (IgG(N)) was conducted using the SARS-CoV-2 IgG II and SARS-CoV-2 IgG assays, respectively [Abbott Diagnostics (Illinois, U.S., RRID:SCR_010477)]. Results were reported in WHO binding antibody units (BAU/ml) based on the mathematical relationship of BAU/ml = 0.142*AU/ml.25 TCR repertoire analysis Sequencing of the CDR3 region of human TCRb chains was performed using the ImmunoSEQ Assay (Adaptive Biotechnologies, Seattle, WA). T-cell response was quantified by measuring clonal breadth (number of spike-specific clones) and depth (cells expressing a specific TCR sequence).26 We used T-detect to set a threshold for TCR response as described previously.22 Statistical analysis To evaluate pre-vaccine immunological correlates of vaccine-induced responses, we combined CyTOF-derived immune cell features with clinical measurements across cancer patients. Two vaccine responses were considered: (1) humoral response, quantified via post-vaccination anti-spike IgG (IgG-S) titers; and (2) cellular response, defined using log-transformed breadth and depth of SARS-CoV-2-specific T-cell repertoires. There was no attrition; data from all patients were included in the analysis. Given the variability in sample collection timing, we normalized IgG-S titers against a densely sampled healthcare worker cohort20 by calculating patient-to-healthcare worker (HCW) titer ratios, where HCW titers were smoothed via a Gaussian-weighted moving average over a 40 d window (σ = 10). This yielded a patient-specific humoral response score independent of time elapsed since vaccination. Cellular response data were available for 53 patients, and hierarchical clustering (Ward method) on ranked IgG-S and TCR scores grouped patients into low, cellular only, and high responders. Associations between immune factors and vaccine responses were evaluated using a covariate-adjusted, bootstrap-based framework. Both immune features and response variables were residualized with respect to age, sex, ethnicity, treatment group, and cancer type (hematologic vs. solid). Covariate-adjusted features were correlated with covariate-adjusted responses using bootstrapped Spearman correlation (5,000 iterations per feature). Bootstrapped correlations were fit to Beta distributions, and two-sided p-values were computed to assess whether the mean correlation significantly differed from zero (p < .1). Given the high dimensionality of the CyTOF dataset and modest cohort size, p < .1 was used to identify response-associated immunologic features suitable for further study. Finally, both covariate-adjusted features and responses were dichotomized using Gaussian Mixture Models (posterior probability > 0.75), and bootstrapped risk ratios and 95% confidence intervals (5,000 iterations) were computed to estimate the relative probability of mounting a vaccine response per feature. Results In this study, we included 66 patients on active treatment, comprising those with B-cell/plasma cell malignancies (n = 44, 66.7%), solid tumors (n = 19, 28.8%), and severe aplastic anemia (n = 3, 4.5%) with median age of 62 (IQR, 54.0–72.0). Treatments within six months before vaccination included B-cell/plasma cell therapy (n = 37, 56.1%), ICI (n = 20, 30.3%), and BMT/immune suppression (n = 9, 13.6%). Patients’ characteristics are summarized in Supplementary Table S1. For each patient, post vaccination humoral response was assessed by measuring the IgG-S titers for cancer patients and the healthy control population of HCW (Figure 1(A)). Humoral response for the cohort was estimated relative to the response of healthcare workers for the corresponding amount of time post vaccination (Figure 1(B)). In a subset of patients (n = 53), we profiled the clonal T-cell repertoire at the same time point corresponding to the IgG-S titer measurement (Figure 1(A)). Patients with combined TCR breadth and depth magnitude above the threshold were classified as having a positive T-cell clonal response (Figure 1(C)). Vaccine response assessment of patients with cancer. (A) Scheme of experimental workflow. (B) Plot shows the log10 IgG-S titers for HCW and cancer cohort relative to time points after primary vaccination (n = 66). (C) Plot shows the scaled values of averaged log-normalized cellular response measures (TCR breadth and depth) across cohort (n = 53). T-detect is used to set a threshold for clonal T-cell response. Red and blue dots represent patients with positive and negative TCR responses, respectively. (D) Cancer patient humoral (left) and scaled cellular response (right) by treatment and cancer groups. (E) Three group hierarchical clustering of patients on percentile-ranked humoral and cellular response, delineating low, cellular only, and high responder annotation. (F) Sex and age of patients (upper row) and treatment and cancer-type breakdown (lower row) by response status. BMT: bone marrow transplantation; HCW: healthcare worker; ICI: immune checkpoint inhibitor; TCR: T-cell receptor. We first evaluated the impact of different treatment modalities on vaccine-induced immune responses. Our analysis revealed that patients who received ICI or underwent BMT/immune suppression exhibited higher humoral and cellular response scores compared to those treated with B-cell/plasma cell – targeted therapies (Figure 1(D)). We next stratified the patient cohort based on the pattern of humoral and cellular responses into the three categories: low, cellular only, and high responders (Figure 1(E)). The majority of low responders were patients with hematologic malignancies receiving B-cell/plasma cell therapies (14/22, 64%). Of note, these patients have also comprised the significant proportion of the cellular only responders (9/11, 82%). Nonetheless, a subset of patients on B-cell/plasma cell therapies (7/20, 35%) was classified as high responders (Figure 1(F)). Moreover, approximately 50% of high responders were patients receiving ICI (heme malignancy: 1/20, 5%; solid tumor: 9/20, 45%). Thus, patients in our cohort exhibited varying levels of vaccine response, likely influenced by their treatment modality, most notably in those receiving B-cell/plasma cell – targeted therapies that demonstrated low vaccine responses. To construct the immunological variables, the peripheral blood samples before primary vaccination were examined for immunophenotyping using a 44-marker panel by CyTOF (Figure 2(A)). To identify the immune cell populations, we primarily clustered the cells based on the expression of lineage-defining markers into T cells (CD3, CD4, and CD8a), B cells (CD19), and non-B/T cells (CD33, CD11b, CD11c, CD1c, CD14, CD16, and CD56). The latter included the cells from myeloid and innate immune compartments. Assessment of immune population frequencies across response categories revealed that B cells (clusters 13, 2, and 22), and T cells (clusters 30, and 6) show at or near significant enrichment differences between response groups (p-value < .05) (Figure 2(B)). To gain a better resolution for every primary cluster, we subclustered them by integrating the associated phenotypic and functional markers (Figure 2(C)). In addition, monocytes and lymphocytes concentrations per patient were extracted from the CBC measurements (Figure 2(D), Supplementary Figure S1). Identification of immunological features in the patient cohort. (A) Heatmap shows the stacked hierarchical-Leiden clustering of CyTOF single-cell data (n = 66) based on normalized expression of restricted lineage marker set with major immune subset annotations including B cell, T cell, and non-B/T cells. (B) Primary cluster log-frequencies (ω) by response-status, showing a Kruskal–Wallis test for differences in means (alpha = 0.05). (C) Normalized expression of sub-clusters driven by integrating additional phenotypic and functional markers into primary clusters. The side bars indicate the absolute number of the corresponding sub-cluster. (D) Mean concentrations of lymphocytes and monocytes for cancer cohort patients in thousands per micro-liter; averaged values for all CBCs spanning period of 60 d before CyTOF collection date to 3 d post-vaccine (first dose). (E) Correlations of lymphocytic and monocytic cluster frequencies with CBC mean concentrations (C), colored by major lineages B, T, and non-B/T cells. (F) Cluster concentrations versus frequencies across cohort. Points are colored by source patient. We further validated that these concentrations correlated with the definition of primary immune clusters of B cells, T cells, and non-B/T cells (Figure 2(E)). Accordingly, we calculated per-patient concentrations of immunological subclusters, which we then used together with monocyte/lymphocyte CBC mean and standard-deviation values, frequencies (Figure 2(F)), and patient-averaged expression of phenotype state markers to create the full set of tested immunologic features. We then identified features and response variables showing significant correlations with either response metric after adjusting for demographic and clinical covariates, including cancer group (i.e. hematologic vs. solid), after which we further quantified risk rations for mounting a response (Supplementary Figure S2). Using a combination of both phenotyping and functional markers, we assigned a label to these immunologic populations (Figure 3(A)) Immunological features associated with humoral and cellular responses. (A) Expression clustermap shows subsets with significant correlations to humoral and cellular response. Total cells count across cohort, humoral and cellular response correlations are indicated in first three columns; average of frequency and concentration correlations used for the respective phenotype. (B,C) Bubble plots display correlations between response variables (humoral, cellular) and immunological features (p < .1). Bubble size represents the –log10(p-value). Star denotes p < .05. Color scale indicates the strength and direction of the correlation (positive or negative). (D) Log-scale median risk-ratios (RR) and 95% bootstrapped confidence intervals for features significantly correlated to seroconversion, and (E) cellular response. The most prominent immunological feature significantly correlating with IgG-S and clonal TCR responses was frequency of naïve (CD27– IgD+) B cells (Figure 3(B,C)). Consistent with this, RR analysis showed that patients with a higher frequency of naïve B cells had an estimated 67% higher probability of mounting a humoral response (RR = 1.67, 95% CI: 0.96–9.09) and a 33% higher probability of mounting a cellular response (RR = 1.33, 95% CI: 0.52–4.60) (Figure 3(D,E), Table S2, S3). Additionally, we identified other B-cell populations, including a subset with elevated Foxp3 and reduced CD19 expression (non-canonical B-reg cells). Of note, expression of chemokine receptors involved in B-cell homing and trafficking (CCR7, CXCR5, and CCR6) on these populations correlated strongly with humoral and cellular responses (Figure 3(B,C)). These findings warrant further investigation into whether these B-cell subsets actively regulate immune cell function in the context of vaccination. In the T-cell compartment, the abundance of naïve CD4+ T cells, and the expression of the co-stimulatory molecule CD27 on these cells positively correlated with the humoral response (Figure 3(B)). Additional T-cell-related immunological features that negatively correlated with antibody and cellular responses included the abundance of CD4+ central memory T cells (TCM) expressing the coinhibitory receptor CD161, CD4+ effector memory T cells (TEM), and senescent CD57+ CD4+ TEM cells expressing PD1, CD8+ TEM cells, as well as expression of coinhibitory receptor CTLA-4 on CD8+ terminally differentiated effector memory T cells (TEMRA) (Figure 3(B,C)). Consistent with this associations, higher frequency of memory T cells was estimated to confer a lower likelihood of mounting humoral and cellular responses (Figure 3(D,E), Table S2, S3). It has been shown that durable spike-specific T-cell responses in healthy individuals following vaccination are mediated by CD4+ and CD8+ memory T cells.14,27 Our data suggest that in cancer patients the presence of expanded preexisting effector memory T cell population with limited proliferation capacity and dysfunctional state may not support the generation of a robust immune response upon new exposure to the SARS-CoV-2 vaccine. Lastly, increased variability in lymphocyte levels (Std Lymphocyte), potentially reflecting treatment effects, was negatively correlated with cellular response (Figure 3(C)). High lymphocyte variability was linked to a 56.7% lower probability of mounting a cellular response (RR = 0.43, 95% CI: 0.08–1.25) in the RR analysis (Figure 3(E), Table S3). In addition, we observed several immune features from innate immune compartments correlated with the immune responses. Strikingly, the concentration of cytolytic (CD56dim CD16+) NK cells, showed a significant positive association with the humoral response (Figure 3(A,B)). Previous studies have shown that SARS-CoV-2 vaccination can activate NK cells and trigger NK cell-mediated antibody-dependent cellular cytotoxicity (ADCC) activity.15,28 Therefore, a high abundance of NK cells and ADCC-mediated elimination of virally infected cells may contribute to an enhanced IgG-S response in cancer patients. Interestingly, the abundance of intermediate monocytes showed a positive correlation, whereas expression of the immune checkpoint inhibitor PD-L1 on classical monocytes was negatively associated with both humoral and TCR responses (Figure 3(B,C)). Accordingly, patients with higher levels of intermediate monocytes were estimated to have a greater probability of generating humoral (RR = 1.36, 95% CI 0.74–6.19) and cellular (RR = 1.56, 95% CI 0.77–6.55) responses (Figure 3(D,E), Table S2, S3). While SARS-CoV-2 vaccination has been reported to promote protective immunity in part through monocyte activation,16 elevated PD-L1 expression on these cells may impair vaccine-induced immune responses by dampening T-cell activation. In addition, the abundance of immunosuppressive MDSCs is also negatively correlated with humoral responses (Figure 3(B)). This is consistent with prior studies showing that MDSC expansion and activation are linked to impaired T-cell function and increased disease severity in COVID-19 patients.29 Discussion We identified distinct pre-vaccine immune features from both the innate and adaptive immune system that were correlated to humoral and cellular responses to SARS-CoV-2 vaccination in cancer patients. Moreover, 96% of patients received mRNA-based vaccines similar to anti-cancer vaccines which are currently (NCT03897881, NCT04161755) under investigations.30 SARS-CoV-2 vaccine response data provided insights into the overall immune fitness, which is highly relevant for guiding subsequent immune-based therapies. Our findings underscore the need to tailor vaccination schedules based on specific immunologic profiles and treatment regimens, highlighting that both immune cell composition and functional state are critical factors for optimal vaccine timing in patients undergoing immune-altering cancer therapies. In line with prior studies, our findings highlight that T-cell – mediated immunity can partially compensate for impaired humoral responses in a subset of patients with hematologic malignancies receiving B-cell – targeted therapies.7,31,32 This cellular response likely contributes to mitigating severe COVID-19 outcomes in this vulnerable population.33 A key limitation of our study was the use of TCR sequencing, which may not fully capture the unique, nonpublic TCR clonotypes, potentially underestimating the breadth of the T-cell response. Furthermore, we acknowledge that the heterogeneity of patients and limited sample size restricted in-depth assessments of response correlates by cancer type and across treatment categories. Future studies with larger, more homogeneous, and treatment-specific cohorts will be essential to validate and expand upon these findings. Moreover, this study was limited to a single post-vaccination timepoint and therefore does not capture longitudinal immune dynamics such as booster-induced responses or durability of immunity. Longitudinal studies with serial sampling will be necessary to define the persistence of vaccine-induced immune responses relative to pre-vaccine immune profile. Our study suggested that the differentiation and functional states of T cells in cancer patients might have important implications for their ability to mount robust antibody and adaptive immune responses following SARS-CoV-2 vaccination; however, the causal relationships remain to be defined. Future studies should focus on profiling the functional status of T cells, for instance, through employing activation-induced marker (AIM) assays or cytokine profiling in a larger, more diverse cohort, which could provide mechanistic validation of these immune correlates, ultimately leveraging vaccine-induced immune responses as a surrogate for immune fitness to personalize treatment strategies for immune-altered patients. Supplementary Material Supplementary Information_Kazerani et al_Revised_Clean.docx Acknowledgments We sincerely appreciate all SeroNet-CORALE participants and study staff. We acknowledge Spatial Molecular Profiling Shared Resource at Cedars-Sinai Medical Center for the support and assistance for CyTOF mass cytometry. This work was also supported by the Biobank and Biostatistics Shared Resources of Cedars-Sinai Cancer. We thank the frontline healthcare workers for their continued dedication to delivering highest quality patient care. Biographies Jane C. Figueiredo is the Program Leader of the Cancer Prevention and Control Program, Professor of Medicine and Director of Community and Population Health Research at the Samuel Oschin Comprehensive Cancer Institute at Cedars-Sinai Medical Center. Her current research program applies a multidisciplinary framework to examine the intersections among biological, environmental, and behavioral/lifestyle aspects of cancer control and survivorship. When SARS-CoV-2 emerged, she pivoted her research focus to studying the immune-inflammatory response to SARS-CoV-2 in immunocompromised cancer patients. Dr. Figueiredo co-leads the U54CA260591 SeroNet-Coronavirus Risk Associations and Longitudinal Evaluation (CORALE) Study. The major goals of the SeroNet-CORALE U54 program are to advance our understanding of SARS-CoV-2 risk, the course of disease, the nature of recovery, and the potential for immunity in patients with cancer undergoing treatment. Several manuscripts have been published, representing major contributions to our understanding of SARS-CoV-2 infection and response to vaccination spanning public health, basic science, and clinical translational impacts. Akil Merchant is Co-Director of the Lymphoma Program, Associate Professor of Medicine, and Director of the Spatial Molecular Profiling Shared Resource (SMPSR) Core Facility at Cedars-Sinai Medical Center. His research focuses on the interaction between cancer and the immune system, with a particular emphasis on predicting responses to immuno-oncology drugs and cellular therapies. Clinically, Dr. Merchant specializes in the treatment of leukemia, lymphoma, and other blood cancers, and leads multiple clinical trials that offer novel therapies for patients with blood cancers. In response to the COVID-19 pandemic, his research expanded to include immune profiling in cancer patients affected by SARS-CoV-2. He co-leads the U54CA260591 SeroNet-Coronavirus Risk Associations and Longitudinal Evaluation (CORALE) Study, which has produced several impactful publications on how cancer immune therapies alter risk for severe COVID-19 and affect B and T cell responses to COVID-19 vaccination in immunocompromised cancer patients. Funding Statement This work was supported by the U.S. National Institutes of Health/National Cancer Institute (NIH/NCI) under grant number [U54CA260591] and by the Biobank and Biostatistics Shared Resource of Cedars-Sinai Cancer. Disclosure statement N.M. holds a consultant or advisory role at Amgen, Kite, Epizyme, TG Therapeutics, ADC Therapeutics, and has research funding from Miltenyi, Teva, and Amgen. J.G. holds a consultant or advisory role at EMD Serono; Elsevier; Exelixis; QED Therapeutics; Natera, Basilea, HalioDx, Eisai, Janssen. O.H. has obtained consulting fees/support meetings/travel/Financial interests in Alkermes, Amgen, Bactonix, Beigene, Bioatla, BMS, Esai, Roche, Genentech, Georgiamune, GigaGen, Grit Bio, GSK, Idera, Immunocore, Incyte, Instilbio, IO Bio, Iovance, Janssen, KSQ, Merck Moderna, Novartis, Obsidian, Pfizer, Regeneron, Sanofi, Seattle Genetics, Tempus, Vial, Zelluna. K.L.R. holds a consultant with honoraria from AstraZeneca, Daiichi Sankyo, Genentech, GlaxoSmithKline, Janssen, Merus, Novartis, Novocure, Pfizer and research funding from Genentech, Blueprint, Daiichi Sankyo, Elevation Oncology, Janssen. J.D. holds a consultant or advisory role Kite Pharma and Morphosys. R.V. is on the Speaker′s Bureau for Amgen, Bristol Myers Squib, Glaxo Smith Klein, Janssen, Karyopharm, and Takeda Pharmaceuticals. A.M. holds a consultant, speaker or advisory role at Novartis, Morphosys, BMS, Genmab, AbbVie, Amgen and Lyell Immunopharma and has received research funding from Amgen and Pfizer. All other authors have declared no potential conflicts of interest. Data availability statement All data generated in this study are available within the article and its Supplementary Information and from the corresponding authors upon reasonable request. De-identified human/patient standardized data will be deposited in ImmPort Shared Data (https://www.immport.org/shared/home) and publicly available as of the date of publication. The original codes are available upon request from the corresponding authors. Supplementary materials Supplemental data for this article can be accessed online at https://doi.org/10.1080/21645515.2025.2599622 References Associated Data Supplementary Materials Supplementary Information_Kazerani et al_Revised_Clean.docx Data Availability Statement All data generated in this study are available within the article and its Supplementary Information and from the corresponding authors upon reasonable request. De-identified human/patient standardized data will be deposited in ImmPort Shared Data (https://www.immport.org/shared/home) and publicly available as of the date of publication. The original codes are available upon request from the corresponding authors.