Search

Single-cell transcriptomics has emerged as a powerful tool for understanding how different cells contribute to disease progression by identifying cell types that change across diseases or conditions. However, detecting changing cell types is challenging due to individual-to-individual and cohort-to-cohort variability and naive approaches based on current computational tools lead to false positive findings. To address this, we propose a computational tool,scDist, based on a mixed-effects model that provides a statistically rigorous and computationally efficient approach for detecting transcriptomic differences. By accurately recapitulating known immune cell relationships and mitigating false positives induced by individual and cohort variation, we demonstrate thatscDistoutperforms current methods in both simulated and real datasets, even with limited sample sizes. Through the analysis of COVID-19 and immunotherapy datasets,scDistuncovers transcriptomic perturbations in dendritic cells, plasmacytoid dendritic cells, and FCER1G+NK cells, that provide new insights into disease mechanisms and treatment responses. As single-cell datasets continue to expand, our faster and statistically rigorous method offers a robust and versatile tool for a wide range of research and clinical applications, enabling the investigation of cellular perturbations with implications for human health and disease.

Although antibodies targeting specific tumor-expressed antigens are the standard of care for some cancers, the identification of cancer-specific targets amenable to antibody binding has remained a bottleneck in development of new therapeutics. To overcome this challenge, we developed a high-throughput platform that allows for the unbiased, simultaneous discovery of antibodies and targets based on phenotypic binding profiles. Applying this platform to ovarian cancer, we identified a wide diversity of cancer targets including receptor tyrosine kinases, adhesion and migration proteins, proteases and proteins regulating angiogenesis in a single round of screening using genomics, flow cytometry, and mass spectrometry. In particular, we identified BCAM as a promising candidate for targeted therapy in high-grade serous ovarian cancers. More generally, this approach provides a rapid and flexible framework to identify cancer targets and antibodies.

Unsupervised clustering of single-cell RNA-sequencing data enables the identification and discovery of distinct cell populations. However, the most widely used clustering algorithms are heuristic and do not formally account for statistical uncertainty. Many popular pipelines use clustering stability methods to assess the algorithms’ output and decide on the number of clusters. However, we find that by not addressing known sources of variability in a statistically rigorous manner, these analyses lead to overconfidence in the discovery of novel cell-types. We extend a previous method for Gaussian data, Significance of Hierarchical Clustering (SHC), to propose a model-based hypothesis testing approach that incorporates significance analysis into the clustering algorithm and permits statistical evaluation of clusters as distinct cell populations. We also adapt this approach to permit statistical assessment on the clusters reported by any algorithm. We benchmarked our approach on real-world datasets against popular clustering workflows, demonstrating improved performance. To show its practical utility, we applied it to the Human Lung Cell Atlas and an atlas of the mouse cerebellar cortex. We identified several cases of over-clustering, leading to false discoveries, as well as under-clustering, resulting in the failure to identify new subpopulations that our method was able to detect.

We introduce mirTarRnaSeq, an R/Bioconductor package for quantitative assessment of miRNA-mRNA relationships within sample cohorts. mirTarRnaSeq is a statistical package to explore predicted or pre-hypothesized miRNA-mRNA relationships following target prediction.We present two use cases applying mirTarRnaSeq. First, to identify miRNA targets, we examined EBV miRNAs for interaction with human and virus transcriptomes of stomach adenocarcinoma. This revealed enrichment of mRNA targets highly expressed in CD105+ endothelial cells, monocytes, CD4+ T cells, NK cells, CD19+ B cells, and CD34 cells. Next, to investigate miRNA-mRNA relationships in SARS-CoV-2 (COVID-19) infection across time, we used paired miRNA and RNA sequenced datasets of SARS-CoV-2 infected lung epithelial cells across three time points (4, 12, and 24 hours post-infection). mirTarRnaSeq identified evidence for human miRNAs targeting cytokine signaling and neutrophil regulation immune pathways from 4 to 24 hours after SARS-CoV-2 infection. Confirming the clinical relevance of these predictions, three of the immune specific mRNA-miRNA relationships identified in human lung epithelial cells after SARS-CoV-2 infection were also observed to be differentially expressed in blood from patients with COVID-19. Overall, mirTarRnaSeq is a robust tool that can address a wide-range of biological questions providing improved prediction of miRNA-mRNA interactions.

Spatial transcriptomics enables spatially resolved gene expression measurements at near single-cell resolution. The detection of genes that are differentially expressed across tissue context for cell types of interest is an essential challenge for dissecting pathological mechanisms. However, changes in cell type composition across space and measurement contamination from other cell types introduce complex statistical challenges. Here, we introduce a statistical method, Generalized Linear Admixture Models for Differential Expression (GLAMDE), that estimates cell type-specific patterns of differential gene expression while accounting for localization of other cell types. By modeling spatial transcriptomics gene expression as an additive mixture, across cell types, of generalized log-linear functions, we provide a unified framework for identifying gene expression changes for a wide-range of relevant contexts: changes due to pathology, anatomical regions, physical proximity to specific cell types, and cellular microenvironment. Our approach enables statistical inference across multiple samples and replicates. We demonstrate, through simulations and validation experiments on Slide-seq and MERFISH data, that our approach accurately identifies cell type-specific differential gene expression with valid uncertainty quantification. Lastly, we apply GLAMDE to characterize spatially-localized tissue changes in the context of disease. In an Alzheimer’s mouse model Slide-seq dataset, we identify plaque-dependent patterns of cellular immune activity. We also find a putative interaction between tumor cells and myeloid immune cells in a Slide-seq tumor dataset. GLAMDE is available within the spacexr R package. D.C. was supported by a Fannie and John Hertz Foundation Fellowship and an NSF Graduate Research Fellowship. This work was supported by an NIH Early Independence Award (DP5, 1DP5OD024583 to F.C.), the NHGRI (R01, R01HG010647 to F.C.), as well as the Burroughs Wellcome Fund, the Searle Scholars Award, and the Merkin Institute to F.C.. R.A.I. was supported by NIH grants R35GM131802 and R01HG005220.

At homeostasis, most Th17 cells are found in the lamina propria of the intestine, where they contribute to tissue homeostasis by inhibiting microbiota from tissue invasion and promote barrier functions. Recent studies in humans and mice have implicated intestinal Th17 cells in extra-intestinal autoimmune diseases, including multiple sclerosis, but the mechanism in which intestinal Th17 cells mediate these dichotomous functions remains unknown. Here, we combined single-cell RNA- and TCR-seq (scRNA/TCR-seq) with fate mapping of tissue Th17 cells to profile over 84,000 tissue Th17 cells and characterize their heterogeneity, plasticity, and migration at homeostasis and during CNS autoimmunity. We discovered a homeostatic Th17 cell population, that is induced by the intestinal microbiota, is present in both lymphoid organs and the intestine, and expresses IL-17. Joint scRNA- and TCR-seq showed that during EAE this homeostatic population gives rise to a pathogenic Th17 cell population, that migrates specifically through the draining lymph nodes and the spleen to the CNS, and highly expresses GM-CSF and IFNγ, but expresses little or low levels of IL-17. This conversion of homeostatic Th17 cells into encephalitogenic Th17 cells depended on IL-23R signaling. Our work characterizes Th17 heterogeneity, plasticity, and migration during homeostasis and CNS autoimmune disease, identifies a mechanism by which intestinal Th17 cells influence the generation of an autoimmune disease-inducing population, and provides further evidence for a direct relationship between the IL-17-producing and GM-CSF- and IFNγ- producing pathogenic T cells.

Comprehensive collections of open reading frame (ORF) deletion mutant strains exist for the budding yeast Saccharomyces cerevisiae. With great prescience, these strains were designed with short molecular bar codes or TAGs that uniquely mark each deletion allele, flanked by shared priming sequences. These features have enabled researchers to handle yeast mutant collections as complex pools of approximately 6000 strains. The presence of any individual mutant within a pool can be assessed indirectly by measuring the relative abundance of its corresponding TAG(s) in genomic DNA prepared from the pool. This is readily accomplished by wholesale polymerase chain reaction (PCR) amplification of the TAGs using fluorescent oligonucleotide primers that recognize the common flanking sequences, followed by hybridization of the labeled PCR products to a TAG oligonucleotide microarray. Here we describe a method-diploid-based synthetic lethality analysis by microarray (dSLAM)-whereby such pools can be manipulated to rapidly construct and assess the fitness of 6000 double-mutant strains in a single experiment. Analysis of double-mutant strains is of growing importance in defining the spectrum of essential cellular functionalities and in understanding how these functionalities interrelate.

TAG, or bar-code, microarrays allow measurement of the oligonucleotide sequences (TAGs) that mark each strain of deletion mutants in the Saccharomyces cerevisiae yeast knockout (YKO) collection. Comparison of genomic DNA from pooled YKO samples allows estimation of relative abundance of TAGs marking each deletion strain. Features of TAG hybridizations create unique challenges for analysis. Analysis is complicated by the presence of two TAGs in most YKO strains and the hybridization behavior of TAGs that may differ in sequence from array probes. The oligonucleotide size of labeled TAGs also results in difficulty with contaminating sequences that cause reduced specificity. We present methods for analysis that approach these unique features of TAG hybridizations.

Although Pseudomonas aeruginosa is the most common bacterial infection in adults with cystic fibrosis and frequently develops resistance to multiple classes of antibiotics, it has not been determined whether patients with multiple antibiotic-resistant Pseudomonas aeruginosa have worse clinical outcomes than patients with more susceptible strains.This study assessed the impact of multiply-resistant P. aeruginosa on lung function, hospitalizations, antibiotic use, lung transplantation and survival in adults with cystic fibrosis.In a cohort study at a university-based adult cystic fibrosis program, 75 consecutive adult cystic fibrosis patients who had P. aeruginosa isolated from sputum cultures were studied over a 4-year period. Outcomes included decline in FEV1, clinic visits, hospitalizations, courses and days of intravenous antibiotics, and lung transplantation. Multiple linear and Poisson regression for repeated measures were used to assess the outcomes.In comparison to patients with susceptible strains, patients with resistant P. aeruginosa had more severe baseline lung disease, more rapid decline in FEV1 (160 ml/year, p = 0.003) and were significantly more likely to undergo lung transplantation (17.6 vs. 0%, p = 0.005).Infection with multiple-antibiotic-resistant P. aeruginosa is associated with accelerated progression of cystic fibrosis, and has important implications for infection control strategies, antibiotic use and lung transplantation.

A time series is a sequence of observations made over time. Examples in public health include daily ozone concentrations, weekly admissions to an emergency department or annual expenditures on health care in the United States. Time series models are used to describe the dependence of the response at each time on predictor variables including covariates and possibly previous values in the series. Time series methods are necessary to account for the correlation among repeated responses over time. This paper gives an overview of time series ideas and methods used in public health research.

Over 10% of the data sets catalogued in the Gene Expression Omnibus Database involve messenger RNA samples that have been pooled prior to hybridization. Pooling affects data quality and inference, but the exact effects are not yet known as pooling has not been systematically studied in the context of microarray experiments. Here we report on the results of an experiment designed to evaluate the utility of pooling and the impact on identifying differentially expressed genes. We find that inference for most genes is not adversely affected by pooling and we recommend that pooling be done when fewer than three arrays are used in each condition. For larger designs, pooling does not significantly improve inferences if few subjects are pooled. The realized benefits in this case do not outweigh the price paid for loss of individual specific information. Pooling is beneficial when many subjects are pooled, provided independent samples contribute to multiple pools.

High density oligonucleotide expression arrays are widely used in many areas of biomedical research. Affymetrix GeneChip arrays are the most popular. In the Affymetrix system, a fair amount of further pre-processing and data reduction occurs following the image processing step. Statistical procedures developed by academic groups have been successful at improving the default algorithms provided by the Affymetrix system. In this paper we present a solution to one of the pre-processing steps, background adjustment, based on a formal statistical framework. Our solution greatly improves the performance of the technology in various practical applications.Affymetrix GeneChip arrays use short oligonucleotides to probe for genes in an RNA sample. Typically each gene will be represented by 11-20 pairs of oligonucleotide probes. The first component of these pairs is referred to as a perfect match probe and is designed to hybridize only with transcripts from the intended gene (specific hybridization). However, hybridization by other sequences (non-specific hybridization) is unavoidable. Furthermore, hybridization strengths are measured by a scanner that introduces optical noise. Therefore, the observed intensities need to be adjusted to give accurate measurements of specific hybridization. One approach to adjusting is to pair each perfect match probe with a mismatch probe that is designed with the intention of measuring non-specific hybridization. The default adjustment, provided as part of the Affymetrix system, is based on the difference between perfect match and mismatch probe intensities. We have found that this approach can be improved via the use of estimators derived from a statistical model that use probe sequence information. The model is based on simple hybridization theory from molecular biology and experiments specifically designed to help develop it.A final step in the pre-processing of these arrays is to combine the 11-20 probe pair intensities,after background adjustment and normalization, for a given gene to define a measure of expression that represents the amount of the corresponding mRNA species. In this paper we illustrate the practical consequences of not adjusting appropriately for the presence of nonspecific hybridization and provide a solution based on our background adjustment procedure. Software that computes our adjustment is available as part of the Bioconductor project (http://www.bioconductor.

Inbred mice have been routinely used in studies of genetic effects that determine behavioral variation due to circadian rhythm. In addition to activity patterns (Act), we aimed to characterize variations in the circadian rhythm of deep-body temperature (T(db)) and heart rate (HR) in a specific genetic model of differential cardiorespiratory control. Radiotelemeters were implanted in C3H/HeJ (C3; n = 11) and C57BL/6J (B6; n = 11) inbred strains. Reciprocal first-generation offspring, B6C3F1/J (B6F1; n = 8) and C3B6F1 (C3F1; n = 3) mice, were included to initiate an evaluation of heritable phenotypes. Mice were housed individually in a facility maintained at 23-24 degrees C, and the light-dark cycle was set at 12-h intervals. In each animal, repeated measurements were obtained at 30-min intervals, and the circadian patterns of Act, T(db), and HR were assessed by novel statistical methods that detailed the periodic function for each strain. During the dark phase, B6 mice demonstrated two distinct peaks in Act and T(db) relative to a single early peak for C3 mice. In contrast to the parental strains, B6F1 and C3F1 mice demonstrated intermediate second peaks in Act and T(db). With respect to HR, the C3 strain demonstrated a significantly (P0.01) greater daily average compared with B6 mice. The circadian rhythm in HR differed significantly from the Act and T(db) patterns in B6 mice (but not in C3 mice); that is, the periodicity in HR for B6 mice preceded the rise and fall in Act and T(db) during both peaks. The B6 phenotype was also observed in F1 mice. In conclusion, these data suggest that the circadian regulation of Act, T(db), and HR vary significantly among C3, B6, and F1 mice. Furthermore, phenotypic differences between C3 and B6 strains can be used to explore the genetic basis for differential circadian regulation of body temperature and HR.

Introduction: Panels of hypermethylated genes have been proposed as diagnostic markers in primary head and neck squamous cell carcinoma using a pharmacological unmasking expression microarray approach. Such experiments are performed in cancer cell lines, mostly with poor relevance when extrapolating to primary cancers. Objective: To overcome this lack of relevance we used two unbiased genome-wide DNA methylation platforms that represent 28K CpG Islands and the promoter regions of 14K-17K genes to identify hypermethylated genes in Oral Squamous Cell Carcinoma (OSCC), which also exhibited downregulated expression in the Gene Expression Omnibus (GEO) repository. Methods: DNA isolated from normal, premalignant lesions and squamous cell carcinoma lesions of the oral cavity was: a) bisulfite treated and hybridized to Infinium arrays; b) enriched with Methylated DNA Immunoprecipitation (MeDIP) and hybridized to Nimblegen 385K tiling arrays. Bioinformatics strategies were used for background correction, array normalization and data analysis of differentially methylated genomic regions between tumor and normal tissue. Genes identified as hypermethylated in cancer with methylation platforms were cross-validated against expression profiles published in the GEO public data repository. Quantitative Methylation Specific PCR (QMSP) was then used to validate candidate genes that were both, hypermethylated and downregulated in cancer. Results: The two unbiased microarray platforms were successful in identifying three novel genes hypermethylated and down regulated in OSCC: MCAM, EDNRB and CALCA. EDNRB, MCAM and CALCA expression was downregulated in HNSCC when compared to normal tissue. A twofold or higher downregulated expression was seen for EDNRB (59%) than for MCAM (55%) and CALCA (55%). The sensitivity of each gene was higher for EDNRB (82%) than for MCAM (60%) and CALCA (76%). The specificity was also higher for EDNRB (86%) than for CALCA (71%), and MCAM (57%). KIF1a had both high sensitivity (82%) and specificity (86%) but there was no expression data in HNSCC available for this gene in the GEO repository. The Receiver Characteristic Operator Curves (ROC) for individual genes revealed area under the curve values (AUC) for: EDNRB (0.88), CALCA (0.85), KIF1a (0.84), and MCAM (0.61). Methylation of three genes was found in 71% of tumor and 0% of normal tissue. Methylation of two genes was found in 82% of tumor and 29% of normal tissue. Conclusion: We successfully demonstrated that unbiased genome-wide methylation arrays paired with publicly available expression data identify panels of relevant hypermethylated genes in primary OSCC. MCAM, KIF1a, EDNRB and CALCA were identified as hypermethylated in OSSC tumor tissue, when compared to normal tissue. These results should be validated in a Phase II Biomarker Development Study. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 2924.

Supplementary files for Chapter 2.