Bioinformatics and Biostatistics of Gastrointestinal Diseases and Geometric Statistics

点击翻译按钮获取中文摘要

This proposal develops novel bioinformatic and biostatistical methodology for metagenomic data. The particular application is for the gastrointestinal diseases ulcerative colitis (UC) and Clostridium difficile (C difficile) infection (CDI). UC is a chronic disease governed by inflammation of the colon while the latter is caused by an aggressive pathogen. Two separate clinical trials were recently conducted and completed with particular attention paid to the clinical response following an experimental fecal microbiota transplantation (FMT). In both clinical trials, FMT provided strong statistical evidence of its efficacy in terms of clinical resolution. A vast quantity of data was collected which includes clinical and demographic data, health survey data and metagenomic data. The clinical response data consisting of resolution following an FMT(s) was recorded in terms of time-to-event. A donor stool provided the material and was paired to the corresponding patient’s pre-FMT stool sample. The donor and pre-FMT patient stool, along with the several subsequent stool samples, were simultaneously sequenced. We note that not all FMTs were successful. Consequently, of particular interest is the development of novel biostatistical methodology for variable selection from the voluminous metagenomic (bioinformatic) covariates that best explains clinical response. More specifically we seek to development a biostatistical methodology to identify what bacteria, in patient and donor colonic microbial systems, play crucial roles in obtaining clinical resolution. The brain-gut axis describes the physiological connection between the gastrointestinal tract and the brain. There is strong evidence that the every day function of the brain, hence the central nervous system, can be altered by modulating the gut microbiota. Although UC and CDI are not neurological diseases some valuable quantifiable information concerning the brain-gut axis could be revealed by the development of variable selection that corresponds the metagenomic covariates to the health survey data. This could provide a methodology for better understanding diseases of the central nervous system such as autism, multiple sclerosis, Parkinson's disease, to name a few. A methodology for the study of the topology of the microbiome will also be pursued. Current metagenomics takes DNA sequences and builds a phylogenetic tree based on phylotyping. If however, we take the unique sequence reads and calculate the distances, using modern computational methods, we can compute the topology of the microbiome. Statistical testing can then be formulated. Typically point cloud data is very noisy and to date there is a paucity in sparse methods in computational algebraic topology. The development of such methods will be pursued with the main application being the bioinformatic data from the gastrointestinal diseases.