Unsupervised Flow Cytometry Analysis Allows for an Accurate Identification of Minimal Residual Disease Assessment in Acute Myeloid Leukemia

MAGIC-DR: An interpretable machine-learning guided approach for acute myeloid leukemia measurable residual disease analysis

Multiparameter flow cytometry is widely used for acute myeloid leukemia minimal residual disease testing (AML MRD) but is time consuming and demands substantial expertise. Machine learning offers potential advancements in accuracy and efficiency, but has yet to be widely adopted for this application. To explore this, we trained single cell XGBoost classifiers from 98 diagnostic AML cell populations and 30 MRD negative samples. Performance was assessed by cross-validation. Predictions were integrated with UMAP as a heatmap parameter for an augmented/interactive AML MRD analysis framework, which was benchmarked against traditional MRD analysis for 25 test cases. The results showed that XGBoost achieved a median AUC of 0.97, effectively distinguishing diverse AML cell populations from normal cells. When integrated with UMAP, the classifiers highlighted MRD populations against the background of normal events. Our pipeline, MAGIC-DR, incorporated classifier predictions and UMAP into flow cytometry standard (FCS) files. This enabled a human-in-the-loop machine learning guided MRD workflow. Validation against conventional analysis for 25 MRD samples showed 100% concordance in myeloid blast detection, with MAGIC-DR also identifying several immature monocytic populations not readily found by conventional analysis. In conclusion, Integrating a supervised classifier with unsupervised dimension reduction offers a robust method for AML MRD analysis that can be seamlessly integrated into conventional workflows. Our approach can support and augment human analysis by highlighting abnormal populations that can be gated on for quantification and further assessment. This has the potential to speed up MRD analysis, and potentially improve detection sensitivity for certain AML immunophenotypes.

Minimal Residual Disease in Acute Myeloid Leukemia: Old and New Concepts

Minimal residual disease (MRD) is of major importance in onco-hematology, particularly in acute myeloid leukemia (AML). MRD measures the amount of leukemia cells remaining in a patient after treatment, and is an essential tool for disease monitoring, relapse prognosis, and guiding treatment decisions. Patients with a negative MRD tend to have superior disease-free and overall survival rates. Considerable effort has been made to standardize MRD practices. A variety of techniques, including flow cytometry and molecular methods, are used to assess MRD, each with distinct strengths and weaknesses. MRD is recognized not only as a predictive biomarker, but also as a prognostic tool and marker of treatment efficacy. Expected advances in MRD assessment encompass molecular techniques such as NGS and digital PCR, as well as optimization strategies such as unsupervised flow cytometry analysis and leukemic stem cell monitoring. At present, there is no perfect method for measuring MRD, and significant advances are expected in the future to fully integrate MRD assessment into the management of AML patients.

Prognostic value of postinduction medullary myeloid recovery by flow cytometry in acute myeloid leukemia

Risk stratification and treatment response evaluation are key features in acute myeloid leukemia (AML) management. Immunophenotypic and molecular approaches all rely on the detection of persisting leukemic cells by measurable residual disease techniques. A new approach is proposed here by assessing medullary myeloid maturation by flow cytometry through a myeloid progenitor ratio (MPR). The normal MPR range was defined using reference normal bone marrows (n = 48). MPR was considered balanced if between 1 and 4 and unbalanced if < 1 or > 4. MPR was retrospectively assessed at baseline and post-induction for 206 newly diagnosed AML patients eligible for intensive treatment from two different French centers. All AML baseline MPR were unbalanced and thus significantly different from normal MPR (p < 0.0001). Patients with an unbalanced MPR after induction had worse 3-year overall survival (OS) (44.4% vs. 80.2%, HR, 2.96; 95% CI, 1.81-4.84, p < 0.0001) and 3-year relapse free survival (RFS) (38.7% vs. 64.4%, HR, 2.11; 95% CI, 1.39-3.18, p < 0.001). In multivariate analysis, postinduction unbalanced MPR was significantly associated with shorter OS and RFS regardless of the European LeukemiaNet 2010 risk stratification or NPM1/FLT3-ITD status. A balanced postinduction MPR conversely conferred favorable outcomes and reflects medullary myeloid recovery.

Computational flow cytometry provides accurate assessment of measurable residual disease in chronic lymphocytic leukaemia

Undetectable measurable residual disease (MRD) is associated with favourable clinical outcomes in chronic lymphocytic leukaemia (CLL). While assessment is commonly performed using multiparameter flow cytometry (MFC), this approach is associated with limitations including user bias and expertise that may not be widely available. Implementation of unsupervised clustering algorithms in the laboratory can address these limitations and have not been previously reported in a systematic quantitative manner. We developed a computational pipeline to assess CLL MRD using FlowSOM. In the training step, a self-organising map was generated with nodes representing the full breadth of normal immature and mature B cells along with disease immunophenotypes. This map was used to detect MRD in multiple validation cohorts containing a total of 456 samples. This included an evaluation of atypical CLL cases and samples collected from two different laboratories. Computational MRD showed high correlation with expert analysis (Pearson's r > 0.99 for typical CLL). Binary classification of typical CLL samples as either MRD positive or negative demonstrated high concordance (>98%). Interestingly, computational MRD detected disease in a small number of atypical CLL cases in which MRD was not detected by expert analysis. These results demonstrate the feasibility and value of automated MFC analysis in a diagnostic laboratory.

From gating to computational flow cytometry: Exploiting artificial intelligence for MRD diagnostics

The era of AI-based methods to improve flow cytometry diagnostics in haematology is now at the beginning. The study by Nguyen and colleagues explored an emerging machine learning approach to assess phenotypic MRD in chronic lymphocytic leukaemia patients, showing that such AI-driven computational analysis may represent a robust and feasible tool for advanced diagnostics of haematological malignancies. Commentary on: Nguyen et al. Computational flow cytometry provides accurate assessment of measurable residual disease in chronic lymphocytic leukaemia. Br J Haematol 2023;202:760-770.

Optimization of monocyte gating to quantify monocyte subsets for the diagnosis of chronic myelomonocytic leukemia

BACKGROUND

The presence of >94% classical monocytes (MO1, CD14++/CD16-) in peripheral blood (PB) has an excellent performance for the diagnosis of chronic myelomonocytic leukemia (CMML). However, the monocyte gating strategy is not well defined. The objective of the study was to compare monocyte gating strategies and propose an optimal one.

METHODS

This is a prospective, single center study assessing monocyte subsets in PB. First, we compared monocyte subsets using 13 monocyte gating strategies in 10 samples. Then we developed our own 10 color tube and tested it on 124 patients (normal white blood cell counts, reactive monocytosis, CMML and a spectrum of other myeloid malignancies). Both conventional and computational (FlowSOM) analyses were used.

RESULTS

Comparing different monocyte gating strategies, small but significant differences in %MO1 and percentually large differences in %MO3 (nonclassical monocytes) were found, suggesting that the monocyte gating strategy can impact monocyte subset quantification. Then, we designed a 10-color tube for this purpose (CD45/CD33/CD14/CD16/CD64/CD86/CD300/CD2/CD66c/CD56) and applied it to 124 patients. This tube allowed proper monocyte gating even in highly abnormal PB. Computational analysis found a higher %MO1 and lower %MO3 compared to conventional analysis. However, differences between conventional and computational analysis in both MO1 and MO3 were globally consistent and only minimal differences were observed when comparing the ranking of patients according to %MO1 or %MO3 obtained with the conventional versus the computational approach.

CONCLUSIONS

The choice of monocyte gating strategy appears relevant for the monocyte subset distribution test. Our 10-color proposal allowed satisfactory monocyte gating even in highly abnormal PB. Computational analysis seems promising to increase reproducibility in monocyte subset quantification.

Artificial Intelligence-Assisted Diagnostic Cytology and Genomic Testing for Hematologic Disorders

Artificial intelligence (AI) is a rapidly evolving field of computer science that involves the development of computational programs that can mimic human intelligence. In particular, machine learning and deep learning models have enabled the identification and grouping of patterns within data, leading to the development of AI systems that have been applied in various areas of hematology, including digital pathology, alpha thalassemia patient screening, cytogenetics, immunophenotyping, and sequencing. These AI-assisted methods have shown promise in improving diagnostic accuracy and efficiency, identifying novel biomarkers, and predicting treatment outcomes. However, limitations such as limited databases, lack of validation and standardization, systematic errors, and bias prevent AI from completely replacing manual diagnosis in hematology. In addition, the processing of large amounts of patient data and personal information by AI poses potential data privacy issues, necessitating the development of regulations to evaluate AI systems and address ethical concerns in clinical AI systems. Nonetheless, with continued research and development, AI has the potential to revolutionize the field of hematology and improve patient outcomes. To fully realize this potential, however, the challenges facing AI in hematology must be addressed and overcome.

A concise review of flow cytometric methods for minimal residual disease assessment in childhood B-cell precursor acute lymphoblastic leukemia

Minimal residual disease refers to a leukemia cell population that is resistant to chemotherapy or radiotherapy and leads to disease relapse. The assessment of MRD is crucial for making an accurate prognosis of the disease and for the choice of optimal treatment strategy. Here, we review the advantages and disadvantages of the available genetic and phenotypic methods and focus on the multiparametric flow cytometry as a promising method with greater sensitivity, speed, and standardization options. In addition, we discuss how the application of automated data analysis outweighs the use of complex combinations of windows and gates in classical analysis, thus eliminating subjective evaluation.

Novel Tools for Diagnosis and Monitoring of AML

In recent years, major advances in the understanding of acute myeloid leukemia (AML) pathogenesis, together with technological progress, have led us into a new era in the diagnosis and follow-up of patients with AML. A combination of immunophenotyping, cytogenetic and molecular studies are required for AML diagnosis, including the use of next-generation sequencing (NGS) gene panels to screen all genetic alterations with diagnostic, prognostic and/or therapeutic value. Regarding AML monitoring, multiparametric flow cytometry and quantitative PCR/RT-PCR are currently the most implemented methodologies for measurable residual disease (MRD) evaluation. Given the limitations of these techniques, there is an urgent need to incorporate new tools for MRD monitoring, such as NGS and digital PCR. This review aims to provide an overview of the different technologies used for AML diagnosis and MRD monitoring and to highlight the limitations and challenges of current versus emerging tools.

Prognostic Impact of Unsupervised Early Assessment of Bulk and Leukemic Stem Cell Measurable Residual Disease in Acute Myeloid Leukemia

PURPOSE

Acute myeloid leukemias (AML) are clonal diseases that develop from leukemic stem cells (LSC) that carry an independent prognostic impact on the initial response to induction chemotherapy, demonstrating the clinical relevance of LSC abundance in AML. In 2018, the European LeukemiaNet published recommendations for the detection of measurable residual disease (Bulk MRD) and suggested the exploration of LSC MRD and the use of multiparametric displays.

EXPERIMENTAL DESIGN

We evaluated the performance of unsupervised clustering for the post-induction assessment of bulk and LSC MRD in 155 patients with AML who received intensive conventional chemotherapy treatment.

RESULTS

The median overall survival (OS) for Bulk+ MRD patients was 16.7 months and was not reached for negative patients (HR, 3.82; P < 0.0001). The median OS of LSC+ MRD patients was 25.0 months and not reached for negative patients (HR, 2.84; P = 0.001). Interestingly, 1-year (y) and 3-y OS were 60% and 39% in Bulk+, 91% and 52% in Bulk-LSC+ and 92% and 88% in Bulk-LSC-.

CONCLUSIONS

In this study, we confirm the prognostic impact of post-induction multiparametric flow cytometry Bulk MRD in patients with AML. Focusing on LSCs, we identified a group of patients with negative Bulk MRD but positive LSC MRD (25.8% of our cohort) with an intermediate prognosis, demonstrating the interest of MRD analysis focusing on leukemic chemoresistant subpopulations.

Clinical application of flow cytometry in patients with unexplained cytopenia and suspected myelodysplastic syndrome: A report of the European LeukemiaNet International MDS-Flow Cytometry Working Group

This article discusses the rationale for inclusion of flow cytometry (FCM) in the diagnostic investigation and evaluation of cytopenias of uncertain origin and suspected myelodysplastic syndromes (MDS) by the European LeukemiaNet international MDS Flow Working Group (ELN iMDS Flow WG). The WHO 2016 classification recognizes that FCM contributes to the diagnosis of MDS and may be useful for prognostication, prediction, and evaluation of response to therapy and follow-up of MDS patients.

Identification of Leukemia-Associated Immunophenotypes by Databaseguided Flow Cytometry Provides a Highly Sensitive and Reproducible Strategy for the Study of Measurable Residual Disease in Acute Myeloblastic Leukemia

Simple Summary

The complete immunophenotypic characterization of acute myeloid leukemia is essential for an accurate diagnosis and follow-up, which is determinant in the course of the disease. In many cases, the only option for the evaluation of minimal residual disease is flow cytometry, so the aim of this study is to develop an automatized multidimensional strategy to identify and characterize LAIPs as well as to detect new emerging aberrances in AML patients during the follow-up. The integrated DFN/LAIP strategy that we propose allows the identification of the most useful markers for minimal residual disease monitoring, improving the sensitivity and specificity of these studies. Furthermore, the use of databases and the automation of the analysis provide the basis for the generation of objective conclusions in minimal residual disease evaluations.

Abstract

Background: Multiparametric Flow Cytometry (MFC) is an essential tool to study the involved cell lineages, the aberrant differentiation/maturation patterns and the expression of aberrant antigens in acute myeloid leukemia (AML). The characterization of leukemia-associated immunophenotypes (LAIPs) at the moment of diagnosis is critical to establish reproducible strategies for the study of measurable residual disease using MFC (MFC-MRD). Methods: In this study, we identify and characterize LAIPs by comparing the leukemic populations of 145 AML patients, using the EuroFlow AML/ MDS MFC panel, with six databases of normal myeloid progenitors (MPCs). Principal component analysis was used to identify and characterize the LAIPs, which were then used to generate individual profiles for MFC-MRD monitoring. Furthermore, we investigated the relationship between the expression patterns of LAIPs and the different subtypes of AML. The MFC-MRD study was performed by identifying residual AML populations that matched with the LAIPs at diagnosis. To further validate this approach, the presence of MRD was also assessed by qPCR (qPCR-MRD). Finally, we studied the association between MFC-MRD and progression-free survival (PFS). Results: The strategy used in this study allowed us to describe more than 300 different LAIPs and facilitated the association of specific phenotypes with certain subtypes of AML. The MFC-MRD monitoring based on LAIPs with good/strong specificity was applicable to virtually all patients and showed a good correlation with qPCR-MRD and PFS. Conclusions: The described methodology provides an objective method to identify and characterize LAIPs. Furthermore, it provides a theoretical basis to develop highly sensitive MFC-MRD strategies.

A Chunking-for-Pooling Strategy for Cytometric Representation Learning for Automatic Hematologic Malignancy Classification

Differentiating types of hematologic malignancies is vital to determine therapeutic strategies for the newly diagnosed patients. Flow cytometry (FC) can be used as diagnostic indicator by measuring the multi-parameter fluorescent markers on thousands of antibody-bound cells, but the manual interpretation of large scale flow cytometry data has long been a time-consuming and complicated task for hematologists and laboratory professionals. Past studies have led to the development of representation learning algorithms to perform sample-level automatic classification. In this work, we propose a chunking-for-pooling strategy to include large-scale FC data into a supervised deep representation learning procedure for automatic hematologic malignancy classification. The use of discriminatively-trained representation learning strategy and the fixed-size chunking and pooling design are key components of this framework. It improves the discriminative power of the FC sample-level embedding and simultaneously addresses the robustness issue due to an inevitable use of down-sampling in conventional distribution based approaches for deriving FC representation. We evaluated our framework on two datasets. Our framework outperformed other baseline methods and achieved 92.3% unweighted average recall (UAR) for four-class recognition on the UPMC dataset and 85.0% UAR for five-class recognition on the hema.to dataset. We further compared the robustness of our proposed framework with that of the traditional downsampling approach. Analysis of the effects of the chunk size and the error cases revealed further insights about different hematologic malignancy characteristics in the FC data.

Mixed Phenotype/Lineage Leukemia: Has Anything Changed for 2021 on Diagnosis, Classification, and Treatment?

Purpose of Review

Recent advances in the small field of the rare mixed phenotype acute leukemias (MPAL) are presented focusing on a better understanding of their pathophysiology and search for better therapeutic approaches.

Recent Findings

Three aspects of respective classification, therapy, and immunophenotype of MPAL are reviewed. New proposals have been made to segregate MPAL subtypes based on their genomic landscape. In parallel, it was found that a large array of therapeutic approaches has been tested in the past few years with increasingly good results. Finally, we explored the use of unsupervised flow cytometry analysis to dissect subtle variations in markers expression to better characterize the variegating aspect of MPALs.

Summary

Genomic and immunophenotypic aspects more clearly link MPAL subtypes with bona fide acute myeloblastic of lymphoblastic leukemias. This is likely to impact therapeutic strategies, towards a better management and outcome.

Unsupervised cluster analysis and subset characterization of abnormal erythropoiesis using the bioinformatic Flow-Self Organizing Maps algorithm

Background

The Flow‐Self Organizing Maps (FlowSOM) artificial intelligence (AI) program, available within the Bioconductor open‐source R‐project, allows for an unsupervised visualization and interpretation of multiparameter flow cytometry (MFC) data.

Methods

Applied to a reference merged file from 11 normal bone marrows (BM) analyzed with an MFC panel targeting erythropoiesis, FlowSOM allowed to identify six subpopulations of erythropoietic precursors (EPs). In order to find out how this program would help in the characterization of abnormalities in erythropoiesis, MFC data from list‐mode files of 16 patients (5 with non‐clonal anemia and 11 with myelodysplastic syndrome [MDS] at diagnosis) were analyzed.

Results

Unsupervised FlowSOM analysis identified 18 additional subsets of EPs not present in the merged normal BM samples. Most of them involved subtle unexpected and previously unreported modifications in CD36 and/or CD71 antigen expression and in side scatter characteristics. Three patterns were observed in MDS patient samples: i) EPs with decreased proliferation and abnormal proliferating precursors, ii) EPs with a normal proliferating fraction and maturation defects in late precursors, and iii) EPs with a reduced erythropoietic fraction but mostly normal patterns suggesting that erythropoiesis was less affected. Additionally, analysis of sequential samples from an MDS patient under treatment showed a decrease of abnormal subsets after azacytidine treatment and near normalization after allogeneic hematopoietic stem‐cell transplantation.

Conclusion

Unsupervised clustering analysis of MFC data discloses subtle alterations in erythropoiesis not detectable by cytology nor FCM supervised analysis. This novel AI analytical approach sheds some new light on the pathophysiology of these conditions.

Technical Aspects of Flow Cytometry-based Measurable Residual Disease Quantification in Acute Myeloid Leukemia: Experience of the European LeukemiaNet MRD Working Party

Measurable residual disease (MRD) quantified by multiparameter flow cytometry (MFC) is a strong and independent prognostic factor in acute myeloid leukemia (AML). However, several technical factors may affect the final read-out of the assay. Experts from the MRD Working Party of the European LeukemiaNet evaluated which aspects are crucial for accurate MFC-MRD measurement. Here, we report on the agreement, obtained via a combination of a cross-sectional questionnaire, live discussions, and a Delphi poll. The recommendations consist of several key issues from bone marrow sampling to final laboratory reporting to ensure quality and reproducibility of results. Furthermore, the experiences were tested by comparing two 8-color MRD panels in multiple laboratories. The results presented here underscore the feasibility and the utility of a harmonized theoretical and practical MFC-MRD assessment and are a next step toward further harmonization.

CD158k and PD-1 expressions define heterogeneous subtypes of Sezary syndrome

SS can be divided into 3 subtypes, each with a different immune environment and response to treatment.

Sezary syndrome (SS) is a rare leukemic form of cutaneous T-cell lymphoma. Diagnosis mainly depends on flow cytometry, but results are not specific enough to be unequivocal. The difficulty in defining a single marker that could characterize Sezary cells may be the consequence of different pathological subtypes. In this study, we used multivariate flow cytometry analyses. We chose to investigate the expression of classical CD3, CD4, CD7, and CD26 and the new association of 2 markers CD158k and PD-1. We performed lymphocyte computational phenotypic analyses during diagnosis and follow-up of patients with SS to define new SS classes and improve the sensitivity of the diagnosis and the follow-up flow cytometry method. Three classes of SS, defined by different immunophenotypic profiles, CD158k⁺ SS, CD158k⁻PD-1⁺ SS, CD158k and PD-1 double-negative SS, showed different CD8⁺ and B-cell environments. Such a study could help to diagnose and define biological markers of susceptibility/resistance to treatment, including immunotherapy.

The Plasmacytoid Dendritic Cell CD123+ Compartment in Acute Leukemia with or without RUNX1 Mutation: High Inter-Patient Variability Disclosed by Immunophenotypic Unsupervised Analysis and Clustering

Plasmacytoid dendritic cells (PDC) constitute a small subset of normal bone marrow (BM) cells but have also been shown to be present, sometimes in large numbers, in several hematological malignancies such as acute myeloid leukemia with RUNX1 mutation, chronic myelomonocytic leukemia or, obviously, blastic plasmacytoid dendritic cell neoplasms. These cells have been reported to display somewhat variable immunophenotypic features in different conditions. However, little is known of their plasticity within individual patients. Using an unsupervised clustering tool (FlowSOM) to re-visit flow cytometry results of seven previously analyzed cases of hematological malignancies (6 acute myeloid leukemia and one chronic myelomonocytic leukemia) with a PDC contingent, we report here on the unexpectedly high variability of PDC subsets. Although five of the studied patients harbored a RUNX1 mutation, no consistent feature of PDCs could be disclosed as associated with this variant. Moreover, the one normal single-node small subset of PDC detected in the merged file of six normal BM could be retrieved in the remission BM samples of three successfully treated patients. This study highlights the capacity of unsupervised flow cytometry analysis to delineate cell subsets not detectable with classical supervised tools.

Unsupervised flow cytometry analysis in hematological malignancies: A new paradigm

Ever since hematopoietic cells became "events" enumerated and characterized in suspension by cell counters or flow cytometers, researchers and engineers have strived to refine the acquisition and display of the electronic signals generated. A large array of solutions was then developed to identify at best the numerous cell subsets that can be delineated, notably among hematopoietic cells. As instruments became more and more stable and robust, the focus moved to analytic software. Almost concomitantly, the capacity increased to use large panels (both with mass and classical cytometry) and to apply artificial intelligence/machine learning for their analysis. The combination of these concepts raised new analytical possibilities, opening an unprecedented field of subtle exploration for many conditions, including hematopoiesis and hematological disorders. In this review, the general concepts and progress achieved in the development of new analytical approaches for exploring high-dimensional data sets at the single-cell level will be described as they appeared over the past few years. A larger and more practical part will detail the various steps that need to be mastered, both in data acquisition and in the preanalytical check of data files. Finally, a step-by-step explanation of the solution in development to combine the Bioconductor clustering algorithm FlowSOM and the popular and widely used software Kaluza® (Beckman Coulter) will be presented. The aim of this review was to point out that the day when these progresses will reach routine hematology laboratories does not seem so far away.