High-dimensional Single-cell Data Research Articles

Interpretable deep learning in single-cell omics.

Single-cell omics technologies have enabled the quantification of molecular profiles in individual cells at an unparalleled resolution. Deep learning, a rapidly evolving sub-field of machine learning, has instilled a significant interest in single-cell omics research due to its remarkable success in analysing heterogeneous high-dimensional single-cell omics data. Nevertheless, the inherent multi-layer nonlinear architecture of deep learning models often makes them 'black boxes' as the reasoning behind predictions is often unknown and not transparent to the user. This has stimulated an increasing body of research for addressing the lack of interpretability in deep learning models, especially in single-cell omics data analyses, where the identification and understanding of molecular regulators are crucial for interpreting model predictions and directing downstream experimental validations. In this work, we introduce the basics of single-cell omics technologies and the concept of interpretable deep learning. This is followed by a review of the recent interpretable deep learning models applied to various single-cell omics research. Lastly, we highlight the current limitations and discuss potential future directions.

Single-Cell Informatics for Tumor Microenvironment and Immunotherapy.

Cancer comprises malignant cells surrounded by the tumor microenvironment (TME), a dynamic ecosystem composed of heterogeneous cell populations that exert unique influences on tumor development. The immune community within the TME plays a substantial role in tumorigenesis and tumor evolution. The innate and adaptive immune cells "talk" to the tumor through ligand-receptor interactions and signaling molecules, forming a complex communication network to influence the cellular and molecular basis of cancer. Such intricate intratumoral immune composition and interactions foster the application of immunotherapies, which empower the immune system against cancer to elicit durable long-term responses in cancer patients. Single-cell technologies have allowed for the dissection and characterization of the TME to an unprecedented level, while recent advancements in bioinformatics tools have expanded the horizon and depth of high-dimensional single-cell data analysis. This review will unravel the intertwined networks between malignancy and immunity, explore the utilization of computational tools for a deeper understanding of tumor-immune communications, and discuss the application of these approaches to aid in diagnosis or treatment decision making in the clinical setting, as well as the current challenges faced by the researchers with their potential future improvements.

Application of Dimension Reduction Methods to High-Dimensional Single-Cell 3D Genomic Contact Data

The volume and complexity of data in various fields, particularly in biology, are increasing exponentially, posing a challenge to existing analytical methods, which often struggle with high-dimensional data such as single-cell Hi-C data. To address this issue, we employ unsupervised methods, specifically Principal Component Analysis (PCA) and t-Distributed Stochastic Neighbor Embedding (t-SNE), to reduce data dimensions for visualization. Furthermore, we assess the information retention of the decomposed components using a Linear Discriminant Analysis (LDA) classifier model. Our findings indicate that these dimensionality reduction techniques effectively capture and present information not readily apparent in the original high-dimensional data, facilitating the visualization and interpretation of complex biological data. The LDA classifier's performance suggests that PCA and t-SNE maintain critical information necessary for accurate classification. In conclusion, our study demonstrates that PCA and t-SNE are powerful tools for visualizing and analyzing high-dimensional biological data, enabling researchers to gain new insights and understandings that are challenging to achieve with traditional approaches.

Macrophage polarization toward M1 phenotype in T cell transfer colitis model

BackgroundT cell transfer colitis model is often used to study the CD4+ T cell functions in the intestine. However, the specific roles of macrophages in colitis remain unclear. In this study, we aimed to evaluate the phenotype and functions of macrophages in the colonic lamina propria (LP) in a colitis model.MethodsColitis was induced in scid mice via the adaptive transfer of CD4+CD45RBhi T cells. Then, flow cytometry was used to determine the number of macrophages in the colonic LP and expression of cytokines in macrophages at the onset of colitis. Moreover, M1/M2 macrophage markers were detected in the colonic LP during colitis development using high-dimensional single-cell data and gating-based analyses. Expression levels of M1 markers in macrophages isolated from the colonic LP were measured using quantitative reverse transcription-polymerase chain reaction. Additionally, macrophages were co-cultured with T cells isolated from the colon to assess colitogenic T cell activation.ResultsInfiltration of macrophages into the colon increased with the development of colitis in the T cell transfer colitis model. M1/M2 macrophage markers were observed in this model, as observed in the colon of patients with inflammatory bowel disease (IBD). Moreover, number of M1 macrophages increased, whereas that of M2 macrophages decreased in the colonic LP during colitis development. M1 macrophages were identified as the main source of inflammatory cytokine production, and colitogenic T cells were activated via interactions with these macrophages.ConclusionsOur findings revealed that macrophages polarized toward the M1 phenotype in LP during colitis development in the T cell transfer colitis model. Therefore, the colitis model is suitable for the evaluation of the efficacy of macrophage-targeted drugs in human IBD treatment. Furthermore, this model can be used to elucidate the in vivo functions of macrophages in the colon of patients with IBD.

Single-cell multi-omics topic embedding reveals cell-type-specific and COVID-19 severity-related immune signatures

The advent of single-cell multi-omics sequencing technology makes it possible for researchers to leverage multiple modalities for individual cells and explore cell heterogeneity. However, the high-dimensional, discrete, and sparse nature of the data make the downstream analysis particularly challenging. Here, we propose an interpretable deep learning method called moETM to perform integrative analysis of high-dimensional single-cell multimodal data. moETM integrates multiple omics data via a product-of-experts in the encoder and employs multiple linear decoders to learn the multi-omics signatures. moETM demonstrates superior performance compared with six state-of-the-art methods on seven publicly available datasets. By applying moETM to the scRNA+ scATAC data, we identified sequence motifs corresponding to the transcription factors regulating immune gene signatures. Applying moETM to CITE-seq data from the COVID-19 patients revealed not only known immune cell-type-specific signatures but also composite multi-omics biomarkers of critical conditions due to COVID-19, thus providing insights from both biological and clinical perspectives.

CosTaL: an accurate and scalable graph-based clustering algorithm for high-dimensional single-cell data analysis.

With the aim of analyzing large-sized multidimensional single-cell datasets, we are describing a method for Cosine-based Tanimoto similarity-refined graph for community detection using Leiden's algorithm (CosTaL). As a graph-based clustering method, CosTaL transforms the cells with high-dimensional features into a weighted k-nearest-neighbor (kNN) graph. The cells are represented by the vertices of the graph, while an edge between two vertices in the graph represents the close relatedness between the two cells. Specifically, CosTaL builds an exact kNN graph using cosine similarity and uses the Tanimoto coefficient as the refining strategy to re-weight the edges in order to improve the effectiveness of clustering. We demonstrate that CosTaL generally achieves equivalent or higher effectiveness scores on seven benchmark cytometry datasets and six single-cell RNA-sequencing datasets using six different evaluation metrics, compared with other state-of-the-art graph-based clustering methods, including PhenoGraph, Scanpy and PARC. As indicated by the combined evaluation metrics, Costal has high efficiency with small datasets and acceptable scalability for large datasets, which is beneficial for large-scale analysis.

P136 T-cell-Specific Biomarkers of Early Treatment Response in Patients with Rheumatoid Arthritis Treated with a Tumour Necrosis Factor Inhibitor

Abstract Background/Aims The use of tumour necrosis factor inhibitors (TNFi) has advanced the clinical management of rheumatoid arthritis (RA). However, only 30-40% of patients who receive initial treatment with a TNFi experience a beneficial response. The discovery of reliable biomarkers of TNFi response is, therefore, a priority. Whilst T-cell dysfunction is documented in RA pathophysiology, T-cell involvement in TNFi treatment response is unclear. We aimed to phenotype peripheral blood-derived T-cells in RA patients over a number of time-points through the first 12-weeks of TNFi treatment; and to apply a cluster-free association testing method developed on single-cell CITE-seq data—co-varying neighbourhood analysis (CNA)—to high-dimensional single-cell CyTOF data. Methods 16 TNFi-naïve RA patients commenced treatment with a TNFi (10 on Adalimumab, 6 on Etanercept) and peripheral blood samples were collected at 6-7 time-points between pre-treatment and 12-weeks. Peripheral blood mononuclear cells were extracted, and stained with a 35-marker CyTOF panel. Treatment response was classified using the EULAR response criteria, and patients were characterised as good, moderate or non-responders. Data were cleaned in FlowJo and analysed using both i) a cluster-based approach, where data were clustered into 12 populations, and linear mixed models were fitted per cluster to test for significant differences in population abundance and ii) a cluster-free approach called CNA where association testing is purely data-driven. Both analyses controlled for age, sex, type of TNFi and TNFi concentration. Results Cluster-based analysis showed at 12-weeks, 1 of the 12 clusters (CD4+CD39+HLA-DR+) significantly increased in abundance in the non-responder cohort compared to the good and moderate responders (adjusted p = 0.037). Interestingly, HLA-DR and CD39 genetic effects have previously been associated with TNFi response. No significant differences were observed with CNA. Conclusion These results demonstrate that significant differences in T-cell abundance can be seen as early as 12-weeks after commencing treatment with a TNFi. Validation of these findings is required before the predictive value of this T-cell population can be tested. Disclosure M.A. Sutcliffe: None. S. Ling: None. A. Banyard: None. L. Rumker: None. S. Viatte: None. A. Morris: None. A. Barton: None. S. Raychaudhuri: None. D. Plant: None.

A cross entropy test allows quantitative statistical comparison of t-SNE and UMAP representations

The advent of high-dimensional single-cell data has necessitated the development of dimensionality-reduction tools. t-Distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP) are the two most frequently used approaches, allowing clear visualization of complex single-cell datasets. Despite the need for quantitative comparison, t-SNE and UMAP have largely remained visualization tools due to the lack of robust statistical approaches. Here, we have derived a statistical test for evaluating the difference between dimensionality-reduced datasets using the Kolmogorov-Smirnov test on the distributions of cross entropy of single cells within each dataset. As the approach uses the inter-relationship of single cells for comparison, the resulting statistic is robust and capable of identifying true biological variation. Further, the test provides a valid distance between single-cell datasets, allowing the organization of multiple samples into a dendrogram for quantitative comparison of complex datasets. These results demonstrate the largely untapped potential of dimensionality-reduction tools for biomedical data analysis beyond visualization.

EPCO-16. ESTIMATING THE DIFFERENTIATION POTENTIAL AND PLASTICITY OF GLIOBLASTOMA CELLS USING STATISTICAL MECHANICS

Abstract Glioblastoma (GBM) is the most common form of adult brain tumors and patients diagnosed with GBM face a very dismal prognosis with a median survival rate of 15 months. The challenges in treating GBM are many, there is a complex cellular and molecular heterogeneity, both between patients and within individual tumors. In a single tumor, individual cancer cells adopt a variety of stem-like, progenitor-like, and more differentiated phenotypes (a.k.a. cell states). Recent advances in single-cell technology have made it possible to detect this diversity of cellular states, but yet we lack a good explanation for how the states arise and if transitions between states can be exploited to therapeutic ends. We here propose a new computational strategy to measure the differentiation potential of GBM cells, based on the Ising models from statistical mechanics. Originally used to model coupled magnets on a lattice, the Ising models can be generalized to represent the probability of a particular configuration of any system of n variables. We show that the models can be fitted to high-dimensional single cell data in a matter of seconds by solving a convex optimization problem, and that the models help explain cell differentiation in terms of minimization of an energy function of the system. In our on-going work we have fitted the model to single cell data from our previous work on state transitions in GBM (Larsson, Dalmo et al, Molecular Systems Biology 2021) and shown that the predicted differentiation potential of cell states is consistent with the cell state hierarchies derived from lineage tracing experiments. Moving forward, we are now exploring how the Ising models can be used to model state transitions in GBM and to predict the effect of targeted therapy on the cell’s differentiation potential.

Full spectrum flow cytometry and mass cytometry: A 32-marker panel comparison.

High-dimensional single-cell data has become an important tool in unraveling the complexity of the immune system and its involvement in homeostasis and a large array of pathologies. As technological tools are developed, researchers are adopting them to answer increasingly complex biological questions. Up until recently, mass cytometry (MC) has been the main technology employed in cytometric assays requiring more than 29 markers. Recently, however, with the introduction of full spectrum flow cytometry (FSFC), it has become possible to break the fluorescence barrier and go beyond 29 fluorescent parameters. In this study, in collaboration with the Stanford Human Immune Monitoring Center (HIMC), we compared five patient samples using an established immune panel developed by the HIMC using their MC platform. Using split samples and the same antibody panel, we were able to demonstrate highly comparable results between the two technologies using multiple data analysis approaches. We report here a direct comparison of two technology platforms (MC and FSFC) using a 32-marker flow cytometric immune monitoring panel that can identify all the previously described and anticipated immune subpopulations defined by this panel.

A Hashing-Based Framework for Enhancing Cluster Delineation of High-Dimensional Single-Cell Profiles.

The online version contains supplementary material available at 10.1007/s43657-022-00056-z.

Characterization of cell-cell communication in autistic brains with single-cell transcriptomes

BackgroundAutism spectrum disorder is a neurodevelopmental disorder, affecting 1–2% of children. Studies have revealed genetic and cellular abnormalities in the brains of affected individuals, leading to both regional and distal cell communication deficits.MethodsRecent application of single-cell technologies, especially single-cell transcriptomics, has significantly expanded our understanding of brain cell heterogeneity and further demonstrated that multiple cell types and brain layers or regions are perturbed in autism. The underlying high-dimensional single-cell data provides opportunities for multilevel computational analysis that collectively can better deconvolute the molecular and cellular events altered in autism. Here, we apply advanced computation and pattern recognition approaches on single-cell RNA-seq data to infer and compare inter-cell-type signaling communications in autism brains and controls.ResultsOur results indicate that at a global level, there are cell-cell communication differences in autism in comparison with controls, largely involving neurons as both signaling senders and receivers, but glia also contribute to the communication disruption. Although the magnitude of changes is moderate, we find that excitatory and inhibitor neurons are involved in multiple intercellular signaling that exhibits increased strengths in autism, such as NRXN and CNTN signaling. Not all genes in the intercellular signaling pathways show differential expression, but genes in the affected pathways are enriched for axon guidance, synapse organization, neuron migration, and other critical cellular functions. Furthermore, those genes are highly connected to and enriched for genes previously associated with autism risks.ConclusionsOverall, our proof-of-principle computational study using single-cell data uncovers key intercellular signaling pathways that are potentially disrupted in the autism brains, suggesting that more studies examining cross-cell type effects can be valuable for understanding autism pathogenesis.

P190 Unsupervised automated clustering of mass cytometry data identifies unique CD4+ T cell subsets in rheumatoid arthritis

Abstract Background/Aims Treating rheumatoid arthritis (RA) continues to rely heavily on a trial-and-error approach. The potential of deep immunophenotyping of peripheral blood remains to be proven in tailoring treatment to an individual’s biological variability. Here, we aim to compare the analysis of single-cell data using an unsupervised automated clustering approach with conventional biaxial gating to identify T cell imbalances RA. Methods Unstimulated and stimulated peripheral blood mononuclear cells from 10 RA and 10 healthy controls (HC) were immunophenotyped by mass cytometry (CyTOF) using a T-cell panel. Firstly, a conventional biaxial gating strategy and standard definitions of Th cell subsets were applied to compare subset frequencies between cases and controls. Secondly, publicly-available unsupervised analytical workflows [1] were used to automatically define cell clusters hypothesis-free. Their differential abundance between RA and HC was assessed with Mixed-effects modelling of Associations of Single Cells (MASC)[2]. Results Applying unsupervised automated clustering algorithms clearly identifies unusual marker combinations and a Th1/Th17 : Th2/Treg imbalance in RA in this small sample size, while conventional analytical techniques fail to do so. Cell clusters, including a classical pro-inflammatory Th1-like (IL-2+ T-bet+) subset and a novel IL-17+ T-bet+ subset were significantly enriched in RA compared to HC (odds ratio 4.9, P = 6.6 x 10-4; odds ratio 5.5, P = 1.1 x 10-2). In contrast, a FoxP3+ IL-2+ HLA-DR+ cluster was amongst those reduced in RA (odds ratio 0.2, P = 2.0 x 10-7). Conclusion Automated clustering techniques are more powerful than conventional analytical methods when analysing high-dimensional single-cell data. A hypothesis-free approach allows identification of novel cell clusters with unexpected marker combinations, such as T-bet and IL-17, which might otherwise be overlooked with biaxial gating.

PhenoComb: a discovery tool to assess complex phenotypes in high-dimensional single-cell datasets.

High-dimensional cytometry assays can simultaneously measure dozens of markers, enabling the investigation of complex phenotypes. However, as manual gating relies on previous biological knowledge, few marker combinations are often assessed. This results in complex phenotypes with the potential for biological relevance being overlooked. Here, we present PhenoComb, an R package that allows agnostic exploration of phenotypes by assessing all combinations of markers. PhenoComb uses signal intensity thresholds to assign markers to discrete states (e.g. negative, low, high) and then counts the number of cells per sample from all possible marker combinations in a memory-safe manner. Time and disk space are the only constraints on the number of markers evaluated. PhenoComb also provides several approaches to perform statistical comparisons, evaluate the relevance of phenotypes and assess the independence of identified phenotypes. PhenoComb allows users to guide analysis by adjusting several function arguments, such as identifying parent populations of interest, filtering of low-frequency populations and defining a maximum complexity of phenotypes to evaluate. We have designed PhenoComb to be compatible with a local computer or server-based use. In testing of PhenoComb's performance on synthetic datasets, computation on 16 markers was completed in the scale of minutes and up to 26 markers in hours. We applied PhenoComb to two publicly available datasets: an HIV flow cytometry dataset (12 markers and 421 samples) and the COVIDome CyTOF dataset (40 markers and 99 samples). In the HIV dataset, PhenoComb identified immune phenotypes associated with HIV seroconversion, including those highlighted in the original publication. In the COVID dataset, we identified several immune phenotypes with altered frequencies in infected individuals relative to healthy individuals. Collectively, PhenoComb represents a powerful discovery tool for agnostically assessing high-dimensional single-cell data. The PhenoComb R package can be downloaded from https://github.com/SciOmicsLab/PhenoComb. Supplementary data are available at Bioinformatics Advances online.

Alignment of single-cell trajectories by tuMap enables high-resolution quantitative comparison of cancer samples.

Single-cell technologies allow characterization of cancer samples as continuous developmental trajectories. Yet, the obtained temporal resolution cannot be leveraged for a comparative analysis due to the large phenotypic heterogeneity existing between patients. Here, we present the tuMap algorithm that exploits high-dimensional single-cell data of cancer samples exhibiting an underlying developmental structure to align them with the healthy development, yielding the tuMap pseudotime axis that allows their systematic, meaningful comparison. We applied tuMap on single-cell mass cytometry data of acute lymphoblastic and myeloid leukemia to reveal associations between the tuMap pseudotime axis and clinics that outperform cellular assignment into developmental populations. Application of the tuMap algorithm on single-cell RNA sequencing data further identified gene signatures of stem cells residing at the very-early parts of the cancer trajectories. The quantitative framework provided by tuMap allows generation of metrics for cancer patients evaluation.

MarkovHC: Markov hierarchical clustering for the topological structure of high-dimensional single-cell omics data with transition pathway and critical point detection.

Clustering cells and depicting the lineage relationship among cell subpopulations are fundamental tasks in single-cell omics studies. However, existing analytical methods face challenges in stratifying cells, tracking cellular trajectories, and identifying critical points of cell transitions. To overcome these, we proposed a novel Markov hierarchical clustering algorithm (MarkovHC), a topological clustering method that leverages the metastability of exponentially perturbed Markov chains for systematically reconstructing the cellular landscape. Briefly, MarkovHC starts with local connectivity and density derived from the input and outputs a hierarchical structure for the data. We firstly benchmarked MarkovHC on five simulated datasets and ten public single-cell datasets with known labels. Then, we used MarkovHC to investigate the multi-level architectures and transition processes during human embryo preimplantation development and gastric cancer procession. MarkovHC found heterogeneous cell states and sub-cell types in lineage-specific progenitor cells and revealed the most possible transition paths and critical points in the cellular processes. These results demonstrated MarkovHC’s effectiveness in facilitating the stratification of cells, identification of cell populations, and characterization of cellular trajectories and critical points.

NK Cell Diversity Influences Outcomes after Cord Blood Transplantation

Natural killer (NK) cells are the first subset of lymphocytes to reconstitute, and likely play an important role in the graft-versus-leukemia response early after allogeneic hematopoietic stem cell transplantation. NK cells are highly diverse, and express an array of activating and inhibiting receptors, which delicately regulate their activation status.To provide a framework to better understand NK cell development and maturation after cord blood transplantation (CBT) and its implication on clinical outcomes, we designed a mass cytometry (also known as cytometry by time of flight) panel of 40 monoclonal antibodies recognizing lineage markers, NK cell receptors and transcription factors to interrogate the NK cell repertoire in prospectively collected peripheral blood (PB) samples from 60 CBT recipients. The resultant high-dimensional single cell data were analyzed by t-Distributed Stochastic Neighbor Embedding (t-SNE) dimension reduction algorithm to generate expression profile graphs.With this mass cytometry platform, we first interrogated and compared the NK cell repertoire in CB (n=10) and adult peripheral blood (PB) (n=20) and used inverse Simpson index to measure the diversity of the NK cell populations. NK cells from CB donors had a significantly lower diversity (median 537, range 311-891) compared to NK cells from healthy adult PB donors (median 2858, range 1228-5630; p<0.0001). Since CMV infection has been shown to drive NK cell maturation, we next asked if CMV infection also alters NK cell diversity by comparing the diversity of the NK cell repertoire in CB (n=10), and in PB from adult healthy donors who were either CMV-seropositive (n=10), or CMV-seronegative (n=10). The diversity of the NK cell repertoire was significantly greater in CMV seropositive healthy individuals (median 4217, range 2672-5630) compared to CMV-seronegative healthy adults (median 2247 , range 2026-2468; P=0.0002) and cord blood donors (median 537, range 311-891; P<0.0001), pointing to an important role for CMV infection in driving NK diversity.Harnessing mass cytometry, we undertook an in-depth examination of the NK repertoire in prospectively collected PB samples from 60 CBT recipients. We first sought to determine if CMV infection early after CBT can drive NK cell maturation and accelerate the acquisition of a diverse repertoire. The median time to CMV reactivation for the cohort was 71 days. CMV reactivation in the first 100 days after CBT (n=25) was associated with significantly greater NK diversity (median 844, range 198-1196) when compared to patients without CMV reactivation (median 405, range 110-889; p=0.0003). Notably, CMV augmented NK diversity mainly in the KIR+ population (median 743, range 207-1500) compared to the KIR- population (median 361, range 92-1038; p=0.007); while in the absence of CMV reactivation, NK cell diversity in the KIR+ and KIR- populations remained similar (median 305, range 102-702 vs. median 332, range 54-830; p=0.98). These data indicate that CMV infection augments NK diversity after CBT, and that this effect is most pronounced within the KIR+ population.We next determined if NK cell diversity can influence outcomes by comparing the diversity of the NK cell repertoire in PB samples collected in the first 100 days after CBT from 15 patients who went on to relapse compared to 45 patients who remained in remission for 2 years. NK cell diversity was significantly higher (median 839, range 416-1069) in patients who remained in remission compared to those who progressed and relapsed (median 446, range 328-925; p=0.03). Moreover, the greatest difference in diversity in the two groups was noted within the KIR+ population (median 788, range 101-1500 for patients in remission vs median 320, range 177-857 for patients who relapsed; p=0.002).In summary, we have used multiparametric mass cytometry to examine expression of 40 NK markers, and to gain insights into the diversity of the NK cell repertoire after CBT. We have shown that CB NK cells have a limited NK cell diversity, which increases in breadth and subsets after transplant. Moreover, we observed that CMV infection reconfigures the NK cell compartment and drives the diversity of the NK cell repertoire, especially within the KIR+ population. These data may provide the basis to rationally design therapeutic strategies that can modulate NK diversity in the settings of infection and transplant. DisclosuresNo relevant conflicts of interest to declare.

CytoPy: An autonomous cytometry analysis framework.

Cytometry analysis has seen a considerable expansion in recent years in the maximum number of parameters that can be acquired in a single experiment. In response to this technological advance there has been an increased effort to develop new computational methodologies for handling high-dimensional single cell data acquired by flow or mass cytometry. Despite the success of numerous algorithms and published packages to replicate and outperform traditional manual analysis, widespread adoption of these techniques has yet to be realised in the field of immunology. Here we present CytoPy, a Python framework for automated analysis of cytometry data that integrates a document-based database for a data-centric and iterative analytical environment. In addition, our algorithm-agnostic design provides a platform for open-source cytometry bioinformatics in the Python ecosystem. We demonstrate the ability of CytoPy to phenotype T cell subsets in whole blood samples even in the presence of significant batch effects due to technical and user variation. The complete analytical pipeline was then used to immunophenotype the local inflammatory infiltrate in individuals with and without acute bacterial infection. CytoPy is open-source and licensed under the MIT license. CytoPy is available at https://github.com/burtonrj/CytoPy, with notebooks accompanying this manuscript (https://github.com/burtonrj/CytoPyManuscript) and software documentation at https://cytopy.readthedocs.io/.

Machine Learning Approaches Identify Genes Containing Spatial Information From Single-Cell Transcriptomics Data.

The development of single-cell sequencing technologies has allowed researchers to gain important new knowledge about the expression profile of genes in thousands of individual cells of a model organism or tissue. A common disadvantage of this technology is the loss of the three-dimensional (3-D) structure of the cells. Consequently, the Dialogue on Reverse Engineering Assessment and Methods (DREAM) organized the Single-Cell Transcriptomics Challenge, in which we participated, with the aim to address the following two problems: (a) to identify the top 60, 40, and 20 genes of the Drosophila melanogaster embryo that contain the most spatial information and (b) to reconstruct the 3-D arrangement of the embryo using information from those genes. We developed two independent techniques, leveraging machine learning models from least absolute shrinkage and selection operator (Lasso) and deep neural networks (NNs), which are applied to high-dimensional single-cell sequencing data in order to accurately identify genes that contain spatial information. Our first technique, Lasso.TopX, utilizes the Lasso and ranking statistics and allows a user to define a specific number of features they are interested in. The NN approach utilizes weak supervision for linear regression to accommodate for uncertain or probabilistic training labels. We show, individually for both techniques, that we are able to identify important, stable, and a user-defined number of genes containing the most spatial information. The results from both techniques achieve high performance when reconstructing spatial information in D. melanogaster and also generalize to zebrafish (Danio rerio). Furthermore, we identified novel D. melanogaster genes that carry important positional information and were not previously suspected. We also show how the indirect use of the full datasets’ information can lead to data leakage and generate bias in overestimating the model’s performance. Lastly, we discuss the applicability of our approaches to other feature selection problems outside the realm of single-cell sequencing and the importance of being able to handle probabilistic training labels. Our source code and detailed documentation are available at https://github.com/TJU-CMC-Org/SingleCell-DREAM/.

Mass Cytometry Analysis of T-Helper Cells.

CD4+ T cells or helper T cells play various roles in the immune response to pathogens, tumors, as well as in asthma, allergy, and autoimmunity. Consequently, there is great interest in the comprehensive investigation of different T helper cell subsets. Here, we use mass cytometry (CyTOF), which is similar to flow cytometry but uses metal ion-tagged antibodies, which are detected using time-of-flight mass spectrometry. CyTOF allows the simultaneous detection of over 40 different antibodies, allowing us to collect high-dimensional single-cell proteomic data on T helper subsets. We use an extensive staining panel with a large number of lineage markers, cytokines, and other functional markers to identify and characterize CD4+ T cell subsets. In this method, human peripheral blood mononuclear cells are stimulated ex vivo with PMA and ionomycin, which activates T cells. Theactivated CD4+ T cells can then be identified as Th1, Th2, or Th17 cells based on their production of IFNγ, IL-4, and IL-17, respectively. Tregs are identified as CD4+CD25+CD127lo. Once Th1, Th2, Th17, and Tregs have been identified, they can be characterized in more detail using the large number of phenotypic and functional markers included in the CyTOF staining panel. Finally, automated and unbiased high-dimensional data analysis tools can be employed to comprehensively characterize T helper cells and discover novel features.