A General Framework for Comparing Embedding Visualizations Across Class-Label Hierarchies.

Projecting high-dimensional vectors into two dimensions for visualization, known as embedding visualization, facilitates perceptual reasoning and interpretation. Comparing multiple embedding visualizations drives decision-making in many domains, but traditional comparison methods are limited by a reliance on direct point correspondences. This requirement precludes comparisons without point correspondences, such as two different datasets of annotated images, and fails to capture meaningful higher-level relationships among point groups. To address these shortcomings, we propose a general framework for comparing embedding visualizations based on shared class labels rather than individual points. Our approach partitions points into regions corresponding to three key class concepts-confusion, neighborhood, and relative size-to characterize intra- and inter-class relationships. Informed by a preliminary user study, we implemented our framework using perceptual neighborhood graphs to defne these regions and introduced metrics to quantify each concept. We demonstrate the generality of our framework with usage scenarios from machine learning and single-cell biology, highlighting our metrics' ability to draw insightful comparisons across label hierarchies. To assess the effectiveness of our approach, we conducted an evaluation study with fve machine learning researchers and six single-cell biologists using an interactive and scalable prototype built with Python, JavaScript, and Rust. Our metrics enable more structured comparisons through visual guidance and increased participants' confdence in their fndings.

Extricating human tumour immune alterations from tissue inflammation.

Immunotherapies have achieved remarkable successes in the treatment of cancer, but major challenges remain1,2. An inherent weakness of current treatment approaches is that therapeutically targeted pathways are not restricted to tumours, but are also found in other tissue microenvironments, complicating treatment3,4. Despite great efforts to define inflammatory processes in the tumour microenvironment, the understanding of tumour-unique immune alterations is limited by a knowledge gap regarding the immune cell populations in inflamed human tissues. Here, in an effort to identify such tumour-enriched immune alterations, we used complementary single-cell analysis approaches to interrogate the immune infiltrate in human head and neck squamous cell carcinomas and site-matched non-malignant, inflamed tissues. Our analysis revealed a large overlap in the composition and phenotype of immune cells in tumour and inflamed tissues. Computational analysis identified tumour-enriched immune cell interactions, one of which yields a large population of regulatory T (Treg) cells that is highly enriched in the tumour and uniquely identified among all haematopoietically-derived cells in blood and tissue by co-expression of ICOS and IL-1 receptor type 1 (IL1R1). We provide evidence that these intratumoural IL1R1+ Treg cells had responded to antigen recently and demonstrate that they are clonally expanded with superior suppressive function compared with IL1R1- Treg cells. In addition to identifying extensive immunological congruence between inflamed tissues and tumours as well as tumour-specific changes with direct disease relevance, our work also provides a blueprint for extricating disease-specific changes from general inflammation-associated patterns.

New interpretable machine-learning method for single-cell data reveals correlates of clinical response to cancer immunotherapy.

We introduce a new method for single-cell cytometry studies, FAUST, which performs unbiased cell population discovery and annotation. FAUST processes experimental data on a per-sample basis and returns biologically interpretable cell phenotypes, making it well suited for the analysis of complex datasets. We provide simulation studies that compare FAUST with existing methodology, exemplifying its strength. We apply FAUST to data from a Merkel cell carcinoma anti-PD-1 trial and discover pre-treatment effector memory T cell correlates of outcome co-expressing PD-1, HLA-DR, and CD28. Using FAUST, we then validate these correlates in cryopreserved peripheral blood mononuclear cell samples from the same study, as well as an independent CyTOF dataset from a published metastatic melanoma trial. Finally, we show how FAUST's phenotypes can be used to perform cross-study data integration in the presence of diverse staining panels. Together, these results establish FAUST as a powerful new approach for unbiased discovery in single-cell cytometry.

A regulatory T cell signature distinguishes the immune landscape of COVID-19 patients from those with other respiratory infections.

Despite recent studies of immunity to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), little is known about how the immune response against SARS-CoV-2 differs from other respiratory infections. We compare the immune signature from hospitalized SARS-CoV-2­infected patients to patients hospitalized prepandemic with influenza or respiratory syncytial virus (RSV). Our in-depth profiling indicates that the immune landscape in SARS-CoV-2 patients is largely similar to flu or RSV patients. Unique to patients infected with SARS-CoV-2 who had the most critical clinical disease were changes in the regulatory T cell (Treg) compartment. A Treg signature including increased frequency, activation status, and migration markers was correlated COVID-19 severity. These findings are relevant as Tregs are considered for therapy to combat the severe inflammation seen in COVID-19 patients. Likewise, having defined the overlapping immune landscapes in SARS-CoV-2, existing knowledge of flu and RSV infections could be leveraged to identify common treatment strategies.

CD101 genetic variants modify regulatory and conventional T cell phenotypes and functions.

We recently reported that the risk of sexually acquired HIV-1 infection is increased significantly by variants in the gene encoding CD101, a protein thought to modify inflammatory responses. Using blood samples from individuals with and without these variants, we demonstrate that CD101 variants modify the prevalence of circulating inflammatory cell types and show that CD101 variants are associated with increased proinflammatory cytokine production by circulating T cells. One category of CD101 variants is associated with a reduced capacity of regulatory T cells to suppress T cell cytokine production, resulting in a reduction in the baseline level of immune quiescence. These data are supported by transcriptomics data revealing alterations in the intrinsic regulation of antiviral pathways and HIV resistance genes in individuals with CD101 variants. Our data support the hypothesis that CD101 contributes to homeostatic regulation of bystander inflammation, with CD101 variants altering heterosexual HIV-1 acquisition by facilitating increased prevalence and altered function of T cell subsets.

A differential regulatory T cell signature distinguishes the immune landscape of COVID-19 hospitalized patients from those hospitalized with other respiratory viral infections.

SARS-CoV-2 infection has caused a lasting global pandemic costing millions of lives and untold additional costs. Understanding the immune response to SARS-CoV-2 has been one of the main challenges in the past year in order to decipher mechanisms of host responses and interpret disease pathogenesis. Comparatively little is known in regard to how the immune response against SARS-CoV-2 differs from other respiratory infections. In our study, we compare the peripheral blood immune signature from SARS-CoV-2 infected patients to patients hospitalized pre-pandemic with Influenza Virus or Respiratory Syncytial Virus (RSV). Our in-depth profiling indicates that the immune landscape in patients infected by SARS-CoV-2 is largely similar to patients hospitalized with Flu or RSV. Similarly, serum cytokine and chemokine expression patterns were largely overlapping. Unique to patients infected with SARS-CoV-2 who had the most critical clinical disease state were changes in the regulatory T cell (Treg) compartment. A Treg signature including increased frequency, activation status, and migration markers was correlated with the severity of COVID-19 disease. These findings are particularly relevant as Tregs are being discussed as a therapy to combat the severe inflammation seen in COVID-19 patients. Likewise, having defined the overlapping immune landscapes in SARS-CoV-2, existing knowledge of Flu and RSV infections could be leveraged to identify common treatment strategies. HIGHLIGHTS: The immune landscapes of hospitalized pre-pandemic RSV and influenza patients are similar to SARS-CoV-2 patientsSerum cytokine and chemokine expression patterns are largely similar between patients hospitalized with respiratory virus infections, including SARS-CoV-2, versus healthy donorsSARS-CoV-2 patients with the most critical disease displayed unique changes in the Treg compartmentadvances in understanding and treating SARS-CoV-2 could be leveraged for other common respiratory infections.

Exponential bounds for the hypergeometric distribution.

We establish exponential bounds for the hypergeometric distribution which include a finite sampling correction factor, but are otherwise analogous to bounds for the binomial distribution due to León and Perron (Statist. Probab. Lett.62 (2003) 345-354) and Talagrand (Ann. Probab.22 (1994) 28-76). We also extend a convex ordering of Kemperman's (Nederl. Akad. Wetensch. Proc. Ser. A76 = Indag. Math.35 (1973) 149-164) for sampling without replacement from populations of real numbers between zero and one: a population of all zeros or ones (and hence yielding a hypergeometric distribution in the upper bound) gives the extreme case.

Virtual-tissue computer simulations define the roles of cell adhesion and proliferation in the onset of kidney cystic disease.

In autosomal dominant polycystic kidney disease (ADPKD), cysts accumulate and progressively impair renal function. Mutations in PKD1 and PKD2 genes are causally linked to ADPKD, but how these mutations drive cell behaviors that underlie ADPKD pathogenesis is unknown. Human ADPKD cysts frequently express cadherin-8 (cad8), and expression of cad8 ectopically in vitro suffices to initiate cystogenesis. To explore cell behavioral mechanisms of cad8-driven cyst initiation, we developed a virtual-tissue computer model. Our simulations predicted that either reduced cell-cell adhesion or reduced contact inhibition of proliferation triggers cyst induction. To reproduce the full range of cyst morphologies observed in vivo, changes in both cell adhesion and proliferation are required. However, only loss-of-adhesion simulations produced morphologies matching in vitro cad8-induced cysts. Conversely, the saccular cysts described by others arise predominantly by decreased contact inhibition, that is, increased proliferation. In vitro experiments confirmed that cell-cell adhesion was reduced and proliferation was increased by ectopic cad8 expression. We conclude that adhesion loss due to cadherin type switching in ADPKD suffices to drive cystogenesis. Thus, control of cadherin type switching provides a new target for therapeutic intervention.

Finite sampling inequalities: an application to two-sample Kolmogorov-Smirnov statistics.

We review a finite-sampling exponential bound due to Serfling and discuss related exponential bounds for the hypergeometric distribution. We then discuss how such bounds motivate some new results for two-sample empirical processes. Our development complements recent results by Wei and Dudley (2012) concerning exponential bounds for two-sided Kolmogorov - Smirnov statistics by giving corresponding results for one-sided statistics with emphasis on "adjusted" inequalities of the type proved originally by Dvoretzky et al. (1956) and by Massart (1990) for one-sample versions of these statistics.