Cluster stability in the analysis of mass cytometry data

Impact of big data analytics on banking: a case study

Purpose

The paper aims to help enterprises gain valuable knowledge about big data implementation in practice and improve their information management ability, as they accumulate experience, to reuse or adapt the proposed method to achieve a sustainable competitive advantage.

Design/methodology/approach

Guided by the theory of technological frames of reference (TFR) and transaction cost theory (TCT), this paper describes a real-world case study in the banking industry to explain how to help enterprises leverage big data analytics for changes. Through close integration with bank's daily operations and strategic planning, the case study shows how the analytics team frame the challenge and analyze the data with two analytic models – customer segmentation (unsupervised) and product affinity prediction (supervised), to initiate the adoption of big data analytics in precise marketing.

Findings

The study reported relevant findings from a longitudinal data analysis and identified some key success factors. First, non-technical factors, for example intuitive analytics results, appropriate evaluation baseline, multiple-wave implementation and selection of marketing channels critically influence big data implementation progress in organizations. Second, a successful campaign also relies on technical factors. For example, the clustering analytics could promote customers' response rates, and the product affinity prediction model could boost efficient transaction and lower time costs.

Originality/value

For theoretical contribution, this paper verified that the outstanding characteristics of online mutual fund platforms brought up by Nagle, Seamans and Tadelis (2010) could not guarantee organizations' competitive advantages from the aspect of TCT.

Unsupervised machine learning reveals key immune cell subsets in COVID-19, rhinovirus infection, and cancer therapy

For an emerging disease like COVID-19, systems immunology tools may quickly identify and quantitatively characterize cells associated with disease progression or clinical response. With repeated sampling, immune monitoring creates a real-time portrait of the cells reacting to a novel virus before disease-specific knowledge and tools are established. However, single cell analysis tools can struggle to reveal rare cells that are under 0.1% of the population. Here, the machine learning workflow Tracking Responders EXpanding (T-REX) was created to identify changes in both rare and common cells across human immune monitoring settings. T-REX identified cells with highly similar phenotypes that localized to hotspots of significant change during rhinovirus and SARS-CoV-2 infections. Specialized MHCII tetramer reagents that mark rhinovirus-specific CD4+ cells were left out during analysis and then used to test whether T-REX identified biologically significant cells. T-REX identified rhinovirus-specific CD4+ T cells based on phenotypically homogeneous cells expanding by ≥95% following infection. T-REX successfully identified hotspots of virus-specific T cells by comparing infection (day 7) to either pre-infection (day 0) or post-infection (day 28) samples. Plotting the direction and degree of change for each individual donor provided a useful summary view and revealed patterns of immune system behavior across immune monitoring settings. For example, the magnitude and direction of change in some COVID-19 patients was comparable to blast crisis acute myeloid leukemia patients undergoing a complete response to chemotherapy. Other COVID-19 patients instead displayed an immune trajectory like that seen in rhinovirus infection or checkpoint inhibitor therapy for melanoma. The T-REX algorithm thus rapidly identifies and characterizes mechanistically significant cells and places emerging diseases into a systems immunology context for comparison to well-studied immune changes.

CyAnno: A semi-automated approach for cell type annotation of mass cytometry datasets

MOTIVATION

For immune system monitoring in large-scale studies at the single-cell resolution using CyTOF, (semi-)automated computational methods are applied for annotating live cells of mixed cell types. Here, we show that the live cell pool can be highly enriched with undefined heterogeneous cells, i.e., 'ungated' cells, and that current semi-automated approaches ignore their modeling resulting in misclassified annotations.

RESULT

We introduce 'CyAnno', a novel semi-automated approach for deconvoluting the unlabeled cytometry dataset based on a machine learning framework utilizing manually gated training data that allows the integrative modeling of 'gated' cell types and the 'ungated' cells. By applying this framework on several CyTOF datasets, we demonstrated that including the 'ungated' cells can lead to a significant increase in the precision of the 'gated' cell types prediction. CyAnno can be used to identify even a single cell type, including rare cells, with higher efficacy than current state-of-the-art semi-automated approaches.

AVAILABILITY

The CyAnno is available as a python script with a user-manual and sample dataset at https://github.com/abbioinfo/CyAnno.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

TYRO3 induces anti-PD-1/PD-L1 therapy resistance by limiting innate immunity and tumoral ferroptosis

Immune checkpoint blockade therapy has demonstrated promising clinical outcomes for multiple cancer types. However, the emergence of resistance as well as inadequate biomarkers for patient stratification have largely limited the clinical benefits. Here, we showed that tumors with high TYRO3 expression exhibited anti-programmed cell death protein 1/programmed death ligand 1 (anti-PD-1/PD-L1) resistance in a syngeneic mouse model and in patients who received anti-PD-1/PD-L1 therapy. Mechanistically, TYRO3 inhibited tumor cell ferroptosis triggered by anti-PD-1/PD-L1 and facilitated the development of a protumor microenvironment by reducing the M1/M2 macrophage ratio, resulting in resistance to anti-PD-1/PD-L1 therapy. Inhibition of TYRO3 promoted tumor ferroptosis and sensitized resistant tumors to anti-PD-1 therapy. Collectively, our findings suggest that TYRO3 could serve as a predictive biomarker for patient selection and a promising therapeutic target to overcome anti-PD-1/PD-L1 resistance.

Current trends in flow cytometry automated data analysis software

Automated flow cytometry (FC) data analysis tools for cell population identification and characterization are increasingly being used in academic, biotechnology, pharmaceutical, and clinical laboratories. The development of these computational methods is designed to overcome reproducibility and process bottleneck issues in manual gating, however, the take-up of these tools remains (anecdotally) low. Here, we performed a comprehensive literature survey of state-of-the-art computational tools typically published by research, clinical, and biomanufacturing laboratories for automated FC data analysis and identified popular tools based on literature citation counts. Dimensionality reduction methods ranked highly, such as generic t-distributed stochastic neighbor embedding (t-SNE) and its initial Matlab-based implementation for cytometry data viSNE. Software with graphical user interfaces also ranked highly, including PhenoGraph, SPADE1, FlowSOM, and Citrus, with unsupervised learning methods outnumbering supervised learning methods, and algorithm type popularity spread across K-Means, hierarchical, density-based, model-based, and other classes of clustering algorithms. Additionally, to illustrate the actual use typically within clinical spaces alongside frequent citations, a survey issued by UK NEQAS Leucocyte Immunophenotyping to identify software usage trends among clinical laboratories was completed. The survey revealed 53% of laboratories have not yet taken up automated cell population identification methods, though among those that have, Infinicyt software is the most frequently identified. Survey respondents considered data output quality to be the most important factor when using automated FC data analysis software, followed by software speed and level of technical support. This review found differences in software usage between biomedical institutions, with tools for discovery, data exploration, and visualization more popular in academia, whereas automated tools for specialized targeted analysis that apply supervised learning methods were more used in clinical settings.

Mass Cytometry Defines Virus-Specific CD4+ T Cells in Influenza Vaccination

The antiviral response to influenza virus is complex and multifaceted, involving many immune cell subsets. There is an urgent need to understand the role of CD4⁺ T cells, which orchestrate an effective antiviral response, to improve vaccine design strategies. In this study, we analyzed PBMCs from human participants immunized with influenza vaccine, using high-dimensional single-cell proteomic immune profiling by mass cytometry. Data were analyzed using a novel clustering algorithm, denoised ragged pruning, to define possible influenza virus–specific clusters of CD4⁺ T cells. Denoised ragged pruning identified six clusters of cells. Among these, one cluster (Cluster 3) was found to increase in abundance following stimulation with influenza virus peptide ex vivo. A separate cluster (Cluster 4) was found to expand in abundance between days 0 and 7 postvaccination, indicating that it is vaccine responsive. We examined the expression profiles of all six clusters to characterize their lineage, functionality, and possible role in the response to influenza vaccine. Clusters 3 and 4 consisted of effector memory cells, with high CD154 expression. Cluster 3 expressed cytokines like IL-2, IFN-γ, and TNF-α, whereas Cluster 4 expressed IL-17. Interestingly, some participants had low abundance of Clusters 3 and 4, whereas others had higher abundance of one of these clusters compared with the other. Taken together, we present an approach for identifying novel influenza virus–reactive CD4⁺ T cell subsets, a method that could help advance understanding of the immune response to influenza, predict responsiveness to vaccines, and aid in better vaccine design.

Unsupervised machine learning reveals key immune cell subsets in COVID-19, rhinovirus infection, and cancer therapy

For an emerging disease like COVID-19, systems immunology tools may quickly identify and quantitatively characterize cells associated with disease progression or clinical response. With repeated sampling, immune monitoring creates a real-time portrait of the cells reacting to a novel virus before disease specific knowledge and tools are established. However, single cell analysis tools can struggle to reveal rare cells that are under 0.1% of the population. Here, the machine learning workflow Tracking Responders Expanding (T-REX) was created to identify changes in both very rare and common cells in diverse human immune monitoring settings. T-REX identified cells that were highly similar in phenotype and localized to hotspots of significant change during rhinovirus and SARS-CoV-2 infections. Specialized reagents used to detect the rhinovirus-specific CD4+ cells, MHCII tetramers, were not used during unsupervised analysis and instead 'left out' to serve as a test of whether T-REX identified biologically significant cells. In the rhinovirus challenge study, T-REX identified virus-specific CD4+ T cells based on these cells being a distinct phenotype that expanded by ≥95% following infection. T-REX successfully identified hotspots containing virus-specific T cells using pairs of samples comparing Day 7 of infection to samples taken either prior to infection (Day 0) or after clearing the infection (Day 28). Mapping pairwise comparisons in samples according to both the direction and degree of change provided a framework to compare systems level immune changes during infectious disease or therapy response. This revealed that the magnitude and direction of systemic immune change in some COVID-19 patients was comparable to that of blast crisis acute myeloid leukemia patients undergoing induction chemotherapy and characterized the identity of the immune cells that changed the most. Other COVID-19 patients instead matched an immune trajectory like that of individuals with rhinovirus infection or melanoma patients receiving checkpoint inhibitor therapy. T-REX analysis of paired blood samples provides an approach to rapidly identify and characterize mechanistically significant cells and to place emerging diseases into a systems immunology context.

Proteomics and bioinformatics analysis of Fasciola hepatica somatic proteome in different growth phases

Fasciola hepatica (F. hepatica) is a well-known zoonotic parasite that is crucial for economic and public health worldwide. Quantitative proteomics studies have been performed on proteins expressed by F. hepatica to investigate the differential expression of proteomes in different growth phases. And the screening of several marker proteins for use as early diagnostic antigens is essential. In this study, high-performance liquid chromatography-tandem mass spectrometry (LC-MS/MS) was conducted to analyze the differences in the expression of F. hepatica somatic proteins in different growth phases. Furthermore, gene ontology (GO) functional annotation, KEGG metabolic pathway, and clustering analyses were also performed. LC-MS/MS identified 629, 2286, 2254, and 2192 proteins in metacercariae, juvenile flukes 28dpi, immature flukes 59dpi, and adult phases, respectively. GO analysis revealed that differentially expressed proteins (DEPs) were mainly involved in transport, localization, metabolism, enzyme regulation, protein folding and binding, and nucleoside and nucleotide binding. The DEPs were enriched in cells, intracellular components, organelles, cytoplasm, vesicles, and membranes. KEGG pathway annotation results showed that the DEPs were involved in metabolism, genetic information processing, environmental information processing, cellular processes, organismal systems, and other processes. These findings provide a theoretical basis for vaccine development and establishing early diagnostic methods in the future.

Unsupervised machine learning reveals risk stratifying glioblastoma tumor cells

A goal of cancer research is to reveal cell subsets linked to continuous clinical outcomes to generate new therapeutic and biomarker hypotheses. We introduce a machine learning algorithm, Risk Assessment Population IDentification (RAPID), that is unsupervised and automated, identifies phenotypically distinct cell populations, and determines whether these populations stratify patient survival. With a pilot mass cytometry dataset of 2 million cells from 28 glioblastomas, RAPID identified tumor cells whose abundance independently and continuously stratified patient survival. Statistical validation within the workflow included repeated runs of stochastic steps and cell subsampling. Biological validation used an orthogonal platform, immunohistochemistry, and a larger cohort of 73 glioblastoma patients to confirm the findings from the pilot cohort. RAPID was also validated to find known risk stratifying cells and features using published data from blood cancer. Thus, RAPID provides an automated, unsupervised approach for finding statistically and biologically significant cells using cytometry data from patient samples.

Key steps and methods in the experimental design and data analysis of highly multi-parametric flow and mass cytometry

High-dimensional, single-cell cell technologies revolutionized the way to study biological systems, and polychromatic flow cytometry (FC) and mass cytometry (MC) are two of the drivers of this revolution. As up to 30-50 dimensions respectively can be measured per single-cell, they allow deep phenotyping combined with cellular functions studies, like cytokine production or protein phosphorylation. In parallel, the bioinformatics field develops algorithms that are able to process incoming data and extract the most useful and meaningful biological information. However, the success of automated analysis tools depends on the generation of high-quality data. In this review we present the most recent FC and MC computational approaches that are used to prepare, process and interpret high-content cytometry data. We also underscore proper experimental design as a key step for obtaining good quality data.

A comparison framework and guideline of clustering methods for mass cytometry data

Background

With the expanding applications of mass cytometry in medical research, a wide variety of clustering methods, both semi-supervised and unsupervised, have been developed for data analysis. Selecting the optimal clustering method can accelerate the identification of meaningful cell populations.

Result

To address this issue, we compared three classes of performance measures, “precision” as external evaluation, “coherence” as internal evaluation, and stability, of nine methods based on six independent benchmark datasets. Seven unsupervised methods (Accense, Xshift, PhenoGraph, FlowSOM, flowMeans, DEPECHE, and kmeans) and two semi-supervised methods (Automated Cell-type Discovery and Classification and linear discriminant analysis (LDA)) are tested on six mass cytometry datasets. We compute and compare all defined performance measures against random subsampling, varying sample sizes, and the number of clusters for each method. LDA reproduces the manual labels most precisely but does not rank top in internal evaluation. PhenoGraph and FlowSOM perform better than other unsupervised tools in precision, coherence, and stability. PhenoGraph and Xshift are more robust when detecting refined sub-clusters, whereas DEPECHE and FlowSOM tend to group similar clusters into meta-clusters. The performances of PhenoGraph, Xshift, and flowMeans are impacted by increased sample size, but FlowSOM is relatively stable as sample size increases.

Conclusion

All the evaluations including precision, coherence, stability, and clustering resolution should be taken into synthetic consideration when choosing an appropriate tool for cytometry data analysis. Thus, we provide decision guidelines based on these characteristics for the general reader to more easily choose the most suitable clustering tools.

Combination anti-CTLA-4 plus anti-PD-1 checkpoint blockade utilizes cellular mechanisms partially distinct from monotherapies

Immune checkpoint blockade therapy has become a critical pillar of cancer therapy. Here, we characterize the cellular mechanisms of monotherapy and combination anti–cytotoxic T lymphocyte antigen-4 plus anti–programmed cell death-1 therapy. Using high-dimensional single-cell profiling, we determine that combination therapy elicits cellular responses that are partially distinct from those induced by either monotherapy. In particular, combination therapy mediates a switch from expansion of phenotypically exhausted cluster of differentiation 8 (CD8) T cells to expansion of activated effector CD8 T cells. In addition, we systematically compare T cell subsets present in matched peripheral blood and tumor tissues to define what aspects of antitumor responses can be observed peripherally. These findings have significant implications for both the cellular mechanisms of action and biomarkers of response to monotherapies and combination therapy.

Immune checkpoint blockade therapy targets T cell-negative costimulatory molecules such as cytotoxic T lymphocyte antigen-4 (CTLA-4) and programmed cell death-1 (PD-1). Combination anti–CTLA-4 and anti–PD-1 blockade therapy has enhanced efficacy, but it remains unclear through what mechanisms such effects are mediated. A critical question is whether combination therapy targets and modulates the same T cell populations as monotherapies. Using a mass cytometry-based systems approach, we comprehensively profiled the response of T cell populations to monotherapy and combination anti–CTLA-4 plus anti–PD-1 therapy in syngeneic murine tumors and clinical samples. Most effects of monotherapies were additive in the context of combination therapy; however, multiple combination therapy-specific effects were observed. Highly phenotypically exhausted cluster of differentiation 8 (CD8) T cells expand in frequency following anti–PD-1 monotherapy but not combination therapy, while activated terminally differentiated effector CD8 T cells expand only following combination therapy. Combination therapy also led to further increased frequency of T helper type 1 (Th1)-like CD4 effector T cells even though anti–PD-1 monotherapy is not sufficient to do so. Mass cytometry analyses of peripheral blood from melanoma patients treated with immune checkpoint blockade therapies similarly revealed mostly additive effects on the frequencies of T cell subsets along with unique modulation of terminally differentiated effector CD8 T cells by combination ipilimumab plus nivolumab therapy. Together, these findings indicate that dual blockade of CTLA-4 and PD-1 therapy is sufficient to induce unique cellular responses compared with either monotherapy.

An R-Derived FlowSOM Process to Analyze Unsupervised Clustering of Normal and Malignant Human Bone Marrow Classical Flow Cytometry Data

Multiparameter flow cytometry (MFC) is a powerful and versatile tool to accurately analyze cell subsets, notably to explore normal and pathological hematopoiesis. Yet, mostly supervised subjective strategies are used to identify cell subsets in this complex tissue. In the past few years, the implementation of mass cytometry and the big data generated have led to a blossoming of new software solutions. Their application to classical MFC in hematology is however still seldom reported. Here, we show how one of these new tools, the FlowSOM R solution, can be applied, together with the Kaluza® software, to a new delineation of hematopoietic subsets in normal human bone marrow (BM). We thus combined the unsupervised discrimination of cell subsets provided by FlowSOM and their expert-driven node-by-node assignment to known or new hematopoietic subsets. We also show how this new tool could modify the MFC exploration of hematological malignancies both at diagnosis (Dg) and follow-up (FU). This can be achieved by direct comparison of merged listmodes of reference normal BM, Dg, and FU samples of a representative acute myeloblastic case tested with the same immunophenotyping panel. This provides an immediate unsupervised evaluation of minimal residual disease. © 2019 International Society for Advancement of Cytometry.

Automated subset identification and characterization pipeline for multidimensional flow and mass cytometry data clustering and visualization

When examining datasets of any dimensionality, researchers frequently aim to identify individual subsets (clusters) of objects within the dataset. The ubiquity of multidimensional data has motivated the replacement of user-guided clustering with fully automated clustering. The fully automated methods are designed to make clustering more accurate, standardized and faster. However, the adoption of these methods is still limited by the lack of intuitive visualization and cluster matching methods that would allow users to readily interpret fully automatically generated clusters. To address these issues, we developed a fully automated subset identification and characterization (SIC) pipeline providing robust cluster matching and data visualization tools for high-dimensional flow/mass cytometry (and other) data. This pipeline automatically (and intuitively) generates two-dimensional representations of high-dimensional datasets that are safe from the curse of dimensionality. This new approach allows more robust and reproducible data analysis,+ facilitating the development of new gold standard practices across laboratories and institutions.

Negative Co-stimulation Constrains T Cell Differentiation by Imposing Boundaries on Possible Cell States

Co-stimulation regulates T cell activation, but it remains unclear whether co-stimulatory pathways also control T cell differentiation. We used mass cytometry to profile T cells generated in the genetic absence of the negative co-stimulatory molecules CTLA-4 and PD-1. Our data indicate that negative co-stimulation constrains the possible cell states that peripheral T cells can acquire. CTLA-4 imposes major boundaries on CD4⁺ T cell phenotypes, whereas PD-1 subtly limits CD8⁺ T cell phenotypes. By computationally reconstructing T cell differentiation paths, we identified protein expression changes that underlied the abnormal phenotypic expansion and pinpointed when lineage choice events occurred during differentiation. Similar alterations in T cell phenotypes were observed after anti-CTLA-4 and anti-PD-1 antibody blockade. These findings implicate negative co-stimulation as a key regulator and determinant of T cell differentiation and suggest that checkpoint blockade might work in part by altering the limits of T cell phenotypes.

Epigenomic-Guided Mass Cytometry Profiling Reveals Disease-Specific Features of Exhausted CD8 T Cells

Exhausted CD8 T (Tex) cells are immunotherapy targets in chronic infection and cancer, but a comprehensive assessment of Tex cell diversity in human disease is lacking. Here, we developed a transcriptomic- and epigenetic-guided mass cytometry approach to define core exhaustion-specific genes and disease-induced changes in Tex cells in HIV and human cancer. Single-cell proteomic profiling identified 9 distinct Tex cell clusters using phenotypic, functional, transcription factor, and inhibitory receptor co-expression patterns. An exhaustion severity metric was developed and integrated with high-dimensional phenotypes to define Tex cell clusters that were present in healthy subjects, common across chronic infection and cancer or enriched in either disease, linked to disease severity, and changed with HIV therapy. Combinatorial patterns of immunotherapy targets on different Tex cell clusters were also defined. This approach and associated datasets present a resource for investigating human Tex cell biology, with implications for immune monitoring and immunomodulation in chronic infections, autoimmunity, and cancer.

Distinct predictive biomarker candidates for response to anti-CTLA-4 and anti-PD-1 immunotherapy in melanoma patients

BACKGROUND

While immune checkpoint blockade has greatly improved clinical outcomes in diseases such as melanoma, there remains a need for predictive biomarkers to determine who will likely benefit most from which therapy. To date, most biomarkers of response have been identified in the tumors themselves. Biomarkers that could be assessed from peripheral blood would be even more desirable, because of ease of access and reproducibility of sampling.

METHODS

We used mass cytometry (CyTOF) to comprehensively profile peripheral blood of melanoma patients, in order to find predictive biomarkers of response to anti-CTLA-4 or anti-PD-1 therapy. Using a panel of ~ 40 surface and intracellular markers, we performed in-depth phenotypic and functional immune profiling to identify potential predictive biomarker candidates.

RESULTS

Immune profiling of baseline peripheral blood samples using CyTOF revealed that anti-CTLA-4 and anti-PD-1 therapies have distinct sets of candidate biomarkers. The distribution of CD4⁺ and CD8⁺ memory/non-memory cells and other memory subsets was different between responders and non-responders to anti-CTLA-4 therapy. In anti-PD-1 (but not anti-CTLA-4) treated patients, we discovered differences in CD69 and MIP-1β expressing NK cells between responders and non-responders. Finally, multivariate analysis was used to develop a model for the prediction of response.

CONCLUSIONS

Our results indicate that anti-CTLA-4 and anti-PD-1 have distinct predictive biomarker candidates. CD4⁺ and CD8⁺ memory T cell subsets play an important role in response to anti-CTLA-4, and are potential biomarker candidates. For anti-PD-1 therapy, NK cell subsets (but not memory T cell subsets) correlated with clinical response to therapy. These functionally active NK cell subsets likely play a critical role in the anti-tumor response triggered by anti-PD-1.

Distinct Cellular Mechanisms Underlie Anti-CTLA-4 and Anti-PD-1 Checkpoint Blockade

Immune-checkpoint blockade is able to achieve durable responses in a subset of patients; however, we lack a satisfying comprehension of the underlying mechanisms of anti-CTLA-4- and anti-PD-1-induced tumor rejection. To address these issues, we utilized mass cytometry to comprehensively profile the effects of checkpoint blockade on tumor immune infiltrates in human melanoma and murine tumor models. These analyses reveal a spectrum of tumor-infiltrating T cell populations that are highly similar between tumor models and indicate that checkpoint blockade targets only specific subsets of tumor-infiltrating T cell populations. Anti-PD-1 predominantly induces the expansion of specific tumor-infiltrating exhausted-like CD8 T cell subsets. In contrast, anti-CTLA-4 induces the expansion of an ICOS⁺ Th1-like CD4 effector population in addition to engaging specific subsets of exhausted-like CD8 T cells. Thus, our findings indicate that anti-CTLA-4 and anti-PD-1 checkpoint-blockade-induced immune responses are driven by distinct cellular mechanisms.