Machine learning can provide insight and support for a variety of decisions. In some areas of medicine, decision-support models are capable of assisting healthcare practitioners in making accurate diagnoses. In this work we explored the application of these techniques to distinguish between two diseases in veterinary medicine; inflammatory bowel disease (IBD) and alimentary lymphoma (ALA). Both disorders are common gastrointestinal (GI) diseases in humans and animals that share very similar clinical and pathological outcomes. Because of these similarities, distinguishing between these two diseases can sometimes be challenging. In order to identify patterns that may help with this differentiation, we retrospectively mined medical records from dogs and cats with histopathologically diagnosed GI diseases. Since the pathology report is the key conveyer of this information in the medical records, our first study focused on its information structure. Other groups have had a similar interest. In 2008, to help insure consistent reporting, the World Small Animal Veterinary Association (WSAVA) GI International Standardization Group proposed standards for recording histopathological findings (HF) from GI biopsy samples. In our work, we extend WSAVA efforts and propose an information model (composed of information structure and terminology mapped to the Systematized Nomenclature of Medicine - Clinical Terms) to be used when recording histopathological diagnoses (HDX, one or more HF from one or more tissues). Next, our aim was to identify free-text HF not currently expressed in the WSAVA format that may provide evidence for distinguishing between IBD and ALA in cats. As part of this work, we hypothesized that WSAVA-based structured reports would have higher classification accuracy of GI disorders in comparison to use of unstructured free-text format. We trained machine learning models in 60 structured, and independently, 60 unstructured reports. Results show that unstructured information-based models using two machine learning algorithms achieved higher accuracy in predicting the diagnosis when compared to the structured information-based models, and some novel free-text features were identified for possible inclusion in the WSAVA-reports. In our third study, we tested the use of machine learning algorithms to differentiate between IBD and ALA using complete blood count and serum chemistry data. Three models (using naïve Bayes, neural networks, and C4.5 decision trees) were trained and tested on laboratory results for 40 Normal, 40 IBD, and 40 ALA cats. Diagnostic models achieved classification sensitivity ranging between 63% and 71% with naïve Bayes and neural networks being superior. These models can provide another non-invasive diagnostic tool to assist with differentiating between IBD and ALA, and between diseased and non-diseased cats. We believe that relying on our information model for histopathological reporting can lead to a more complete, consistent, and computable knowledgebase in which machine learning algorithms can more efficiently identify these and other disease patterns. / Ph. D. / Computational models play an important role in supporting the decision making process. In some areas of medicine, decision-support models assist healthcare practitioners to make accurate diagnoses. In this work, we explored the application of computational techniques to distinguish between two diseases; inflammatory bowel disease (IBD) and alimentary lymphoma (ALA). These are common gastrointestinal (GI) diseases in humans and animals that share very similar laboratory findings. Because of these similarities, distinguishing between these two diseases can sometimes be challenging. In order to identify patterns that may help with this differentiation, we mined medical records from dogs and cats diagnosed with GI diseases. Since the pathology report is a key source of information for the diagnosis of these two diseases, in our first study we focused on its information structure. Others with similar interest have also examined reports of this type. In 2008, a group proposed standards for recording histopathological findings (HF) from GI biopsy samples. In our work, we extend the group’s efforts and propose an information model (composed of information structure and terminology) to be used when recording histopathological diagnoses (HDX, one or more HF from one or more tissues). Next, our aim was to identify free-text HF not currently expressed in the standardization group’s format that may provide evidence for distinguishing between IBD and ALA in cats. We trained computational models in 60 structured, and independently, 60 unstructured reports. Results show that unstructured information-based models using two computational models achieved higher accuracy in predicting the diagnosis when compared to the structured information-based models. As a result, novel free text features, which improved the performance of the structured reports, were identified. In our third study, we tested the use of computational models to differentiate between IBD and ALA using routine laboratory results. Three models were trained and tested on laboratory results from 40 Normal, 40 IBD, and 40 ALA cats. Diagnostic models achieved classification sensitivity ranging between 63% and 71%. These models can provide another noninvasive diagnostic tool to assist with differentiating between IBD and ALA, and between diseased and non-diseased cats. We believe that relying on our information model for histopathological reporting can lead to a more complete, consistent, and computable knowledgebase for the identification of these two diseases.

京都大学 / 新制・論文博士 / 博士(医学) / 乙第13635号 / 論医博第2326号 / 新制||医||1074(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 齋藤 潤, 教授 森信 暁雄, 教授 松田 文彦 / 学位規則第4条第2項該当 / Doctor of Medical Science / Kyoto University / DFAM

Intestinal homeostasis is a delicate balance between suppression of immune responses against innocuous antigens and stimulation of immune responses against pathogens. Type I interferon (IFN-1) cytokines have both immunostimulatory and immunomodulatory effects. Colon mononuclear phagocytes (MP) constitutively produced IFN-1 in a TRIFdependent manner. We explored the function of endogenous IFN-1 in the colon using the T cell adoptive transfer model of colitis. Transfer of CD4⁺CD45RB^hi naïve T cells from wild type (WT) or IFNAR subunit 1 knockout (IFNAR1^-/-) mice into RAG^-/- hosts resulted in similar onset and severity of colitis. In contrast, RAG^-/- x IFNAR1^-/- double knockout (DKO) mice developed accelerated severe colitis compared to RAG^-/- hosts when transferred WT CD4⁺CD45RB^hi T cells. Although WT or IFNAR1^-/- regulatory T (Treg) cells equally prevented disease caused by CD45RB^hi naïve T cells, WT Treg cells co-transferred with naïve CD4⁺ T cells into DKO recipients failed to expand or maintain Foxp3 expression and gained effector functions in the colon. IFNAR signaling on host hematopoietic cells inhibited T cell-mediated colitis, but not innate colitis. MPs isolated from the colon lamina propria (cLP) required IFNAR signaling for the production of the anti-inflammatory cytokines, IL-10, IL-27, and IL-1RA, but not for the production of classic pro-inflammatory cytokines. IFN-1-dependent secretion of IL-1RA was particularly important in inhibiting the migration of inflammatory DCs with potent T cell proliferative capacity from the cLP to the mesenteric lymph nodes. Finally, preliminary results suggested that IFN-1 may shape the commensal microbiota, but is not essential for controlling specific colitis-inducing bacteria.

Introduction: Early growth response 1 (EGR1) is a zinc-finger transcription factor involved in the regulation of cell growth. It can act as either a tumour suppressor or a tumour promoter with a role in the induction of apoptosis in cancer cells by various pathways and is likely to play a role in colorectal cancer (CRC). EGR1 also appears to play a significant role in inflammatory pathways, therefore a possible role in Inflammatory Bowel Disease (IBD) is hypothesised. Patients with IBD have a greater risk of developing CRC, which is increased with duration of symptoms and severity of inflammation and dysplasia. The aim of this study is to determine whether EGR1 is differentially expressed in diseased colon tissue and to investigate novel EGR1-protein interactions in CRC cell lines. Methods: The relative EGR1 expression in CRC cell lines and in normal mucosa and tumours of colorectal cancer patients was determined by qRT-PCR. IBD patient samples were also examined for differential EGR1 expression levels by qRT-PCR, before and after stimulation with inflammatory mediators. Statistical analysis of the data was performed using ‘R’ statistical package, with the mixed-model ANOVA. Statistical significance was set at < 0.05. The genotype of three EGR1 variants was determined in the samples using PCR and sequencing, and the methylation status of regions of the EGR1 promoter was determined using bisulfite sequencing. A yeasttwo hybrid screen was conducted with EGR1 as bait, and screened against a SW480 CRC cell line library. Interesting novel interactions were investigated using immunocytochemistry and immunoprecipitation, as was the novel interaction between EGR1 and NOD2 and between EGR1 and components of the cytoskeleton. Results: Investigation into the relative EGR1 mRNA expression in CRC has shown that there is differential expression of EGR1 between matched normal mucosa and tumour. EGR1 expression is decreased in IBD patients compared with healthy controls. Induction of EGR1 by inflammatory stimuli also appears to be aberrant in these patients. The differential expression of EGR1 was not associated with aberrant methylation of a large region of the EGR1 promoter in either the CRC or IBD patients or with the genotype of EGR1 variants. EGR1 localises to both the cytoplasm and the nucleus in CRC cell lines and this study demonstrate interactions with the IBD susceptibility protein NOD2 and with components of the cyotskeleton. A yeast-two hybrid screen conducted with EGR1 as bait using a CRC cell line library has identified several other novel protein interactions of EGR1 in CRC cell lines. Conclusion: EGR1 is differentially expressed in both CRC and IBD, and in the case of IBD shows aberrant activity, suggesting that EGR1 may play a role in both colorectal diseases. EGR1 interacts with the IBD protein NOD2, and components of the cytoskeleton in CRC cells. Several novel protein interactions with EGR1 have been identified and warrant further study.

Fecal microbiota transplantation (FMT) is the process by which fecal suspension from a healthy individual is transferred into the gastrointestinal tract of another individual in an attempt to cure certain diseases. This transplantation process has been accredited as being a potential remedy for a growing number of diseases that have been associated with gut microbial imbalances. Interest in FMT has largely been driven by the science community’s increasing interest in the gut microbiome and its role in potentially regulating a multitude of different functions and processes within the human body. One disease that has been found to respond exceptionally well to FMT treatments is Clostridium difficile infection (CDI). However, while FMT has demonstrated high cure rates for CDI, this transplantation process is no panacea. In fact, the results from FMT treatments on other diseases, such as Inflammatory Bowel Disease (IBD), have not been as impressive as CDI’s. This review will examine the existing literature surrounding FMT usage on IBD and will propose a series of experiments and studies needed to truly test the safety and efficacy of FMT for IBD patients. This review will also reference current literature documenting FMT treatments for CDI as a comparative tool for investigating if this form of bacteriotherapy is indeed a viable therapeutic option for treating IBD.

Introduction: Inflammatory pouch complications, including pouchitis, chronic pouchitis (CP) and a Crohn’s disease-like phenotype (CDL) of the pouch following ileal pouch-anal anastomosis (IPAA), are relatively common, and arise via unknown mechanisms. The phenotypic similarities between pouch inflammation and inflammatory bowel disease (IBD) suggest there may be common pathways involved in both disorders. The aim of this thesis is to investigate the serological, genetic and microbial factors contributing to the development of pouch inflammation in a large, well characterized patient cohort. Methods: Subjects with IPAA were recruited, and clinical and demographic information was obtained through medical chart review and patient questionnaire, allowing patients to be grouped based on post-surgical phenotype. Blood and tissue was collected for genetic, serological and microbial analyses. Anti-microbial antibodies were detected using enzyme-linked immunosorbent assay (ELISA), genotyping was carried out using the Illumina Goldengate custom SNP assay and Sequenome iPLEX platform, and tissue-associated microbial communities were assessed using 454 pyrosequencing. Results: Among our cohort, smoking was associated with CDL (P=0.003) and Ashkenazi Jewish heritage with CP (P<0.008). NOD2insC (rs2066847) (P=7.4x10-5), anti-CBir1 (P<0.0001) and ASCA (IgG) (P=0.03) were significantly associated with inflammatory pouch outcomes. Additional SNPs in NOX3, DAGLB, and NCF4 were also marginally associated with pouch outcome. A multi-variable risk model combining clinical, serologic and genetic markers was constructed and could differentiate between chronic pouch inflammatory phenotypes and no pouchitis. Genus level microbial analysis demonstrated that several organisms (Bacteroides, Parabacteroides, Blautia and Moryella) were detected less frequently among the inflammatory outcome groups (P<0.05). These associations remained significant even following adjustment for antibiotic use, smoking, country of birth and gender. Conclusions: CD-associated anti-microbial antibodies and genetic markers are associated with chronic inflammatory pouch phenotypes. Additionally, changes in the composition of the pouch associated microbiome are associated with inflammation. These observations suggest that similar mechanisms may be involved in non-surgical IBD and pouchitis.

Background and objectives: Physical fitness and stress resilience may influence the risk of gastrointestinal (GI) disease. High physical fitness level may reduce levels of systemic inflammation while psychosocial stress exposure can increase inflammation levels and intestinal permeability. The main objectives are to evaluate if poorer physical fitness and stress resilience in adolescence are associated with a raised risk of inflammatory bowel disease (IBD), peptic ulcer disease (PUD) and GI infections in adulthood and to assess evidence of causality. Materials and methods: Swedish registers provided information on a cohort of approximately 250,000 men who underwent military conscription assessments in late adolescence (1969 –1976) with follow-up until December 2009 (up to age 57 years). Cox regression evaluated the associations of physical fitness and stress resilience in adolescence with subsequent GI disease risk in adulthood. Results and conclusions: IBD: Poor physical fitness was associated with an increased risk of IBD. The association may be explained (in part) by prodromal disease activity reducing exercise capacity and therefore fitness. Low stress resilience was associated with an increased risk of receiving an IBD diagnosis. Stress may not be an important cause of IBD but may increase the likelihood of conversion from subclinical to symptomatic disease. PUD: Low stress resilience was associated with an increased risk of PUD. This may be explained by a combination of physiological and behavioural mechanisms that increase susceptibility to H. pylori infections and other risk factors. GI infections: Low stress resilience was associated with a reduced risk of GI infections, including enteric infections rather than the hypothesised increased risk.

Ulcerative colitis (UC) and Crohn’s disease (CD) are the two common forms of inflammatory bowel disease (IBD). They share similar clinical and demographic features as well as harbor key differences in tissue damage and prognosis. Previous studies indicated that they contributed to the increased rick to Colorectal cancer (CRC). However, whether UC and CD share inflammatory signatures still remains controversial. In addition, no inflammatory signatures have been reported on CRC. To answer these questions, a comprehensive study has been conducted on collected microarray datasets. Our analysis suggests that although CD and UC share common inflammatory pathways, they also present difference. Especially, CD patients are likely to have type I response, while UC patients are inclined to undergo type II response. Pathway enrichment analysis on CRC uncovered two potential CRC-specific inflammatory pathways.

INTRODUCTION: The techniques currently in practice to diagnose and assess interval disease activity in patients with inflammatory bowel disease (IBD) are costly and invasive. Physicians typically rely on information derived from a combination of endoscopic, radiologic, and histologic studies to diagnose and determine the extent and severity of the two most common forms of IBD, Crohn disease (CD) and ulcerative colitis (UC). The development of noninvasive methods of assessing response to therapy is of increasing importance to pediatric healthcare providers. Previous studies have demonstrated that serum and fecal biomarkers are reliable measures of inflammation in the gastrointestinal tract. However, existing biomarkers are non-specific and their levels can be elevated in the context of either acute and chronic inflammation (IBD) or infection. As such, further studies are required to develop newer and novel biomarkers that have greater specificity for use in the diagnosis and interval assessment in children and adults with IBD. OBJECTIVES: The goal of this study is to further assess the relationship between biomarkers in the stool and serum of patients with IBD that are being treated with the anti-TNF therapy, infliximab (Remicade). To accomplish this, we will assess the changes in serum and fecal biomarker levels over the course of treatment and correlate the changes in fecal and serum biomarker levels with clinical, biochemical, and endoscopic outcome variables. METHODS: We conducted a prospective longitudinal cohort study in pediatric patients with IBD receiving long-term immunosuppressive therapy with Remicade. Pediatric patients diagnosed with either CD or UC who receive Remicade at Boston Children’s Hospital were recruited. Patients were drawn from subsets of patients that were either naïve to Remicade, had received Remicade for less than 6 months, or had received Remicade for more than one year at the time of enrollment. We collected longitudinal data over the course of their first 6 consecutive infusions following enrollment, including blood and stool samples, disease activity indexes, as well as a patient-reported outcome measure (IMPACT-III Questionnaire) at each infusion session. RESULTS: A total of 33 patients with IBD who fit our eligibility criteria and provided informed consent were enrolled to date. Of these, 20 had a CD diagnosis and 13 had a UC diagnosis. We collected baseline serum, fecal, and IMPACT-III score data and followed enrolled patients over the course of subsequent infusions. Mean baseline fecal ASCA levels from 8 CD and 6 UC patients were 0.08±0.021 OD and 0.02±0.0015 OD, respectively. At baseline, serum lactoferrin (p<0.10), calprotectin (p<0.10), ESR (p<0.05), and CRP (p<0.10) were significantly higher among CD patients. CONCLUSION: Our data demonstrate the potential for serum and fecal biomarkers to evaluate therapeutic response to Remicade. Completion of study enrollment and data collection will be necessary to determine if individual or combinations of fecal and serum biomarkers yield the most robust measures for use in the diagnosis and interval assessment of children and adults with IBD.

Iron deficiency anemia (IDA) is a common extra-intestinal manifestation of inflammatory bowel disease (IBD). It is particularly common in the pediatric population, with 40-60% of pediatric patients with IBD meeting criteria for anemia (Aljomah et al, 2018). A number of studies have examined the use of both IV and oral iron treatments to treat anemia in IBD, but few have examined the safety and efficacy of these treatments in children and how they impact a patient’s health-related quality of life. We conducted a prospective cohort study of 79 pediatric patients admitted to Boston Children’s Hospital for management of their IBD. 48 of these patients received IV iron, 13 received oral iron, and 12 were not treated with either. Treatment with IV iron resulted in a statistically significant increase in hemoglobin of 1.75g/dL+/-1.4g/dL (mean +/- SD) from admission to the time of their first post-discharge visit (p=0.0001). Prescription of oral iron at the time of discharge did not result in a significant increase in hemoglobin over the same interval (p=0.481). Though there was a positive trend, IV iron treatment did not result in a significant change in health-related quality of life (HRQOL) measurements as measured by the IMPACT-III questionnaire (p=0.06). Our study suggests that IDA is common in patients admitted for management of their IBD, IV iron is more efficacious in raising hemoglobin, and that larger studies will be needed to more fully demonstrate the impact of effective iron repletion on overall quality of life in these patients.