Cytometry Part A最新文献

Although flow cytometry produces reliable results, the data processing from gating to fingerprinting is prone to subjective bias. Here, we integrated autogating with Automated Machine Learning in flow cytometry to enhance the classification of bacterial phenotypes. We analyzed six bacterial strains prevalent in the soil and groundwater— Bacillus subtilis , Burkholderia thailandensis , Corynebacterium glutamicum , Escherichia coli , Pseudomonas putida , and Pseudomonas stutzeri . Using the H2O-AutoML framework, we applied gradient-boosting machine (GBM) models to classify bacteria across different metabolic phases. Our results demonstrated an overall classification accuracy of 82.34% for GBM. Notably, accuracy varied across metabolic phases, with the highest observed during the late log (88.06%), lag (88.43%), and early log phases (89.37%), whereas the stationary phase showed a slightly lower accuracy of 80.73%. P. stutzeri exhibited consistently high sensitivity and specificity across all the phases, which indicated that it was the most distinctly identifiable strain. In contrast, E. coli showed low sensitivity, particularly in the stationary phase, which indicated challenges in its classification. Overall, this study with incorporating autogating and the AutoML framework, substantially reduces subjective biases and enhances the reproducibility and accuracy of microbial classification. Our methodology offers a robust framework for microbial classification in flow cytometric analysis, paving the way for more precise and comprehensive analyses of microbial ecology.

Ligand-induced formation of signaling platforms composed of homo- and/or heterodimers of receptor tyrosine kinases is considered essential for their activation and consequential contribution to the progression of many cancers. Epidermal Growth Factor Receptor (EGFR) acts as a signal receiver upon EGF binding and produces mitogenic input for many cells also through receptor-heterodimerization with its ligandless partner, Human Epidermal growth factor Receptor 2 (HER2). Ligand-driven transactivation is a key step leading to changes in the cell surface pattern of EGFR and HER2; their interaction plays a key role in various malignancies, especially when HER2 molecules are overexpressed. Our clinically relevant model system is the SK-BR-3 breast tumor cell line, overexpressing HER2 and moderately expressing EGFR. This cell line shows significant dependency on EGF-driven HER2 signaling. We studied changes in the interaction between EGFR and HER2 in the cell membrane upon EGF binding, applying various biophysical approaches with different time scales. Changes in molecular proximity were characterized by fluorescence lifetime imaging microscopy (FLIM) techniques assessing Förster resonance energy transfer (FRET), which confirmed the ligand-enhanced interaction of EGFR and HER2, followed by an increase in HER2 homoassociation. EGF binding and transactivation were reflected in the phosphorylation of both receptor types as well. At the same time, superresolution Airyscan microscopy and fluorescence correlation and cross-correlation spectroscopy (FCS/FCCS), sensitive to changes in the size of stationary and diffusing aggregates, respectively, have revealed cyclic increases in the aggregation and stable co-diffusion of membrane-localized HER2, possibly caused by internalization and recycling, eventually leading to a new equilibrium. Such dynamic fluctuation of receptor interaction may open a window for the binding of therapeutic antibodies that are aimed at inhibiting heterodimerization, such as pertuzumab. The complementary array of state-of-the-art imaging cytometry approaches thus demonstrates a spatiotemporal pattern of spontaneous and induced receptor aggregation states that could provide mechanistic insights into the potential success of targeted therapies directed at the HER family of receptor tyrosine kinases.

Representing and quantifying Measurable Residual Disease (MRD) in Acute Myeloid Leukemia (AML), a type of cancer that affects the blood and bone marrow, is essential in the prognosis and follow-up of AML patients. As traditional cytological analysis cannot detect leukemia cells below 5%, the analysis of flow cytometry datasets is expected to provide more reliable results. In this paper, we explore statistical learning methods based on optimal transport (OT) to achieve a relevant low-dimensional representation of multi-patient flow cytometry measurements (FCM) datasets considered as high-dimensional probability distributions. Using the framework of OT, we justify the use of the K-means algorithm for dimensionality reduction of multiple large-scale point clouds through mean measure quantization by merging all the data into a single point cloud. After this quantization step, the visualization of the intra-and inter-patient FCM variability is carried out by embedding low-dimensional quantized probability measures into a linear space using either Wasserstein Principal Component Analysis (PCA) through linearized OT or log-ratio PCA of compositional data. Using a publicly available FCM dataset and a FCM dataset from Bordeaux University Hospital, we demonstrate the benefits of our approach over the popular kernel mean embedding technique for statistical learning from multiple high-dimensional probability distributions. We also highlight the usefulness of our methodology for low-dimensional projection and clustering patient measurements according to their level of MRD in AML from FCM. In particular, our OT-based approach allows a relevant and informative two-dimensional representation of the results of the FlowSom algorithm, a state-of-the-art method for the detection of MRD in AML using multi-patient FCM.

Automation and the increased number of measurable parameters in flow cytometry (FCM) have strongly increased the volume and complexity of phenotyping immune cell populations. Despite numerous automated gating methods for FCM analysis, their adoption in routine practice remains challenging due to accessibility barriers for users and potential model failures. Here, we propose a user-centered solution that combines elements of supervised machine learning (SML), rapid application development (RAD), systematic quality assurance guided by structured argumentation, and uncertainty estimation to address these challenges. We implement a data-driven model for event classification and use RAD to generate software prototypes, allowing FCM users to apply the model for automated gating. Considering concepts for structured argumentation from assurance cases (ACs), we derived and justified quality analyses that inform users about the quality of the model. We propose guiding the model operation phase using uncertainty estimation to provide users with a clear understanding of the model's confidence in its predictions. We aim to overcome barriers to the routine application of automated gating and contribute to more reliable and efficient FCM data analysis. Our approach is based on the application of phenotyping for human immune cells. We encourage future research to investigate the potential of SML, ACs, and uncertainty estimation to address dependability of data-driven models (DDMs) supporting diagnostic decision making in the medical domain, including FCM in clinical applications and highly regulated areas such as pharmaceutical research.

T cells play a pivotal role in the immune system's response to various conditions. They are activated by antigen-presenting cells (APCs) via T-cell surface receptors, resulting in cytokine production and T-cell proliferation. These interactions occur through the formation of immunological synapses. The advent of imaging flow cytometry has enabled detailed statistical analyses of these cellular interactions. However, the dynamics of T-cell receptors in response to in vitro stimulation are yet to receive attention, despite it being a crucial aspect of understanding T-cell behavior. In this article, we explore the responses of T cells to in vitro antibody-based stimulation without APCs. Specifically, we established a Th1 cell clone, subjected it to a combination of centrifugation-induced mechanical stress and anti–human CD3 and anti–human CD28 antibody stimulation as the in vitro antibody-based stimulation, and captured and analyzed bright-field and fluorescence images of single cells various hours after stimulation using an imaging flow cytometer. Our results indicate distinct temporal dynamics of CD3 and CD28. Notably, CD3 and CD28 relocated on the T-cell surface immediately after stimulation, with CD3 receptors dispersing after 3.5 h, whereas CD28 remained clustered for 7.5 h. These receptor morphological changes precede cytokine production, suggesting their potential as early indicators of T-cell activation.

Cytometry is a single cell, high-dimensional, high-throughput technique that is being applied across a range of disciplines. However, many elements alongside the data acquisition process might give rise to technical variation in the dataset, called batch effects. CytoNorm is a normalization algorithm for batch effect removal in cytometry data that was originally published in 2020 and has been applied on a variety of datasets since then. Here, we present CytoNorm 2.0, discussing new, illustrative use cases to increase the applicability of the algorithm and showcasing new visualizations that enable thorough quality control and understanding of the normalization process. We explain how CytoNorm can be used without the need for technical replicates or controls, show how the goal distribution can be tailored toward the experimental design and we elaborate on the choice of markers for CytoNorm's internal FlowSOM clustering step.