One primary reason that makes single-cell RNA-seq analysis challenging is dropouts, where the data only captures a small fraction of the transcriptome of each cell. Almost all computational algorithms developed for single-cell RNA-seq adopted gene selection, dimension reduction or imputation to address the dropouts. Here, an opposite view is explored. Instead of treating dropouts as a problem to be fixed, we embrace it as a useful signal. We represent the dropout pattern by binarizing single-cell RNA-seq count data, and present a co-occurrence clustering algorithm to cluster cells based on the dropout pattern. We demonstrate in multiple published datasets that the binary dropout pattern is as informative as the quantitative expression of highly variable genes for the purpose of identifying cell types. We expect that recognizing the utility of dropouts provides an alternative direction for developing computational algorithms for single-cell RNA-seq analysis.

Background: Human gut microbiota are important for human health and have been regarded as a “forgotten organ”, whose variation is closely linked with various factors, such as host genetics, diet, pathological conditions and external environment. The diversity of human gut microbiota has been correlated with aging, which was characterized by different abundance of bacteria in various age groups. In the literature, most of the previous studies of age-related gut microbiota changes focused on individual species in the gut community with supervised methods. Here, we aimed to examine the underlying aging progression of the human gut microbial community from an unsupervised perspective. Results: We obtained raw 16S rRNA sequencing data of subjects ranging from newborns to centenarians from a previous study, and summarized the data into a relative abundance matrix of genera in all the samples. Without using the age information of samples, we applied an unsupervised algorithm to recapitulate the underlying aging progression of microbial community from hosts in different age groups and identify genera associated to this progression. Literature review of these identified genera indicated that for individuals with advanced ages, some beneficial genera are lost while some genera related with inflammation and cancer increase. Conclusions: The multivariate unsupervised analysis here revealed the existence of a continuous aging progression of human gut microbiota along with the host aging process. The identified genera associated to this aging process are meaningful for designing probiotics to maintain the gut microbiota to resemble a young age, which hopefully will lead to positive impact on human health, especially for individuals in advanced age groups.

Background: DNA methylation is a form of epigenetic modification that has been shown to play a significant role in gene regulation. In cancer, DNA methylation plays an important role by regulating the expression of oncogenes. The role of DNA methylation in the onset and progression of various cancer types is now being elucidated as more large-scale data become available. The Cancer Genome Atlas (TCGA) provides a wealth of information for the analysis of various molecular aspects of cancer genetics. Gene expression data and DNA methylation data from TCGA have been used for a variety of studies. A traditional understanding of the effects of DNA methylation on gene expression has linked methylation of CpG sites in the gene promoter region with the decrease in gene expression. Recent studies have begun to expand this traditional role of DNA methylation. Results: Here we present a pan-cancer analysis of correlation patterns between CpG methylation and gene expression. Using matching patient data from TCGA, 33 cancer-specific correlations were calculated for each CpG site and the expression level of its corresponding gene. These correlations were used to identify patterns on a per-site basis as well as patterns of methylation across the gene body. Using these identified patterns, we found genes that contain conflicting methylation signals beyond the commonly accepted association between the promoter region methylation and silencing of gene expression. Beyond gene body methylation in whole, we examined individual CpG sites and show that, even in the same gene body, some sites can have a contradictory effect on gene expression in cancers. Conclusions: We observed that within promoter regions there was a substantial amount of positive correlation between methylation and gene expression, which contradicts the commonly accepted association. We observed that the correlation between CpG methylation and gene expression does not exhibit in a tissue-specific manner, suggesting that the effects of methylation on gene expression are largely tissue independent. The analysis of correlation associated with the location of the CpG site in the gene body has led to the identification of several different methylation patterns that affect gene expression, and several examples of methylation activating gene expression were observed. Distinctly opposing or conflicting effects were seen in close proximity on the gene body, where negative and positive correlations were seen at the neighboring CpG sites.

Current management of childhood leukemia is tailored based on disease risk determined by clinical features at presentation. Whether properties of the host immune response impact disease risk and outcome is not known. Here, we combine mass cytometry, single cell genomics, and functional studies to characterize the BM immune environment in children with B cell acute lymphoblastic leukemia and acute myelogenous leukemia at presentation. T cells in leukemia marrow demonstrate evidence of chronic immune activation and exhaustion/dysfunction, with attrition of naive T cells and TCF1 ⁺ stem-like memory T cells and accumulation of terminally differentiated effector T cells. Marrow-infiltrating NK cells also exhibit evidence of dysfunction, particularly in myeloid leukemia. Properties of immune cells identified distinct immune phenotype–based clusters correlating with disease risk in acute lymphoblastic leukemia. High-risk immune signatures were associated with expression of stem-like genes on tumor cells. These data provide a comprehensive assessment of the immune landscape of childhood leukemias and identify targets potentially amenable to therapeutic intervention. These studies also suggest that properties of the host response with depletion of naive T cells and accumulation of terminal-effector T cells may contribute to the biologic basis of disease risk. Properties of immune microenvironment identified here may also impact optimal application of immune therapies, including T cell–redirection approaches in childhood leukemia.

Local biomaterial-mediated delivery of PD-L1 induces alloislet graft survival and function in a murine model of type 1 diabetes.

Background: Machine learning has been utilized to predict cancer drug response from multi-omics data generated from sensitivities of cancer cell lines to different therapeutic compounds. Here, we build machine learning models using gene expression data from patients’ primary tumor tissues to predict whether a patient will respond positively or negatively to two chemotherapeutics: 5-Fluorouracil and Gemcitabine. Results: We focused on 5-Fluorouracil and Gemcitabine because based on our exclusion criteria, they provide the largest numbers of patients within TCGA. Normalized gene expression data were clustered and used as the input features for the study. We used matching clinical trial data to ascertain the response of these patients via multiple classification methods. Multiple clustering and classification methods were compared for prediction accuracy of drug response. Clara and random forest were found to be the best clustering and classification methods, respectively. The results show our models predict with up to 86% accuracy; despite the study’s limitation of sample size. We also found the genes most informative for predicting drug response were enriched in well-known cancer signaling pathways and highlighted their potential significance in chemotherapy prognosis. Conclusions: Primary tumor gene expression is a good predictor of cancer drug response. Investment in larger datasets containing both patient gene expression and drug response is needed to support future work of machine learning models. Ultimately, such predictive models may aid oncologists with making critical treatment decisions.