| Summary: | Barrett's esophagus is a pre-malignant lesion leading to esophageal adeno-carcinoma, where presence of Barrett's carries a seven fold increase in risk of developing cancer. The histology of Barrett's esophagus is diverse, containing squamous, gastric, and intestinal features, and the origin of the lesion is still under considerable debate. Professor Lu's research group aims to define the molecular features underlying the Barrett's lesion and elucidate the functional origin of the lesion. Because pathogens and commensals have previously been reported to be important in other cancers of the upper gastro-intestinal tract, we also aim to explore the role of microbes and viruses in Barrett's esophagus. We sequenced 3069 single cell RNA samples from endoscopic biopsies of the Barrett's lesion and relevant normal tissue in the upper gastro-intestinal tract and positive and negative controls. In order to draw meaningful conclusions from these single cell RNA data concerning the functional origin of Barrett's, we needed to identify robust relationships between single cell observations across tissue types. However, technical variance is a major confounding factor in single-cell RNA sequencing and could distort any such observed relationships. Therefore, I developed BEARscc, a tool that evaluates cluster robustness to noise by simulating experiment-specific technical replicates. I demonstrate that the tool improves the unsupervised classification of cells and aids the interpretation of single-cell RNA-seq experiments. I apply BEARscc to our single cell RNA-sequencing cohort and find striking similarity between the esophagus submucosal glandular ducts and a subset of Barrett's cells with undi erentiated characteristics. Using BEARscc, I also provide a robust description of the substructure of normal esophageal, gastric, and duodenal tissues. In order to characterize any microbial or viral presence in the microenvironment of Barrett's, I developed a pipeline, poshTitan, to identify pathogens and commensals in next generation sequencing data. I demonstrate that the poshTitan identifies pathogens with known associations in nasopharyngeal squamous cell carcinoma whole tissue RNA-sequencing and gastric adenocarcinoma whole genome sequencing data. I apply the poshTitan to whole tissue and single cell RNA-sequencing samples of Barrett's esophagus and normal controls and find no clear pattern of pathogen or commensal association with disease.
|
|---|
