Development of Histological and Gene Expression-Based Mathematical Models to Objectively Diagnose Coeliac Disease Severity

Abstract

<p><b>Introduction:</b> Coeliac Disease (CD) is a chronic disorder which results from the interplay of both genetic and environmental factors and is characterised as an autoimmune enteropathy triggered by several antigenic epitopes generated from the digestion of gluten, a protein found in many major grains. The disease can manifest at any point in life in individuals on a gluten-containing diet and usually presents primarily with malabsorptive symptomology. The current form of treatment for CD is a strict and lifelong gluten-free diet (GFD). CD is currently diagnosed by both serological and histological assessments, with the latter being the most conclusive current test for the condition. There is debate however as to the accuracy of histological assessment, especially for equivocal or mild CD patients due to the subjective nature of the histological assessment.</p><p><b> Hypothesis and Aims:</b> As the immunological/physiological pathways and processes of CD have not been fully elucidated, the major focus of this thesis was to use quantitative histological and gene expression data derived from duodenal biopsies of CD patients and control patients to further investigate these mechanisms. It was hypothesised that a more accurate classification of CD tissue pathology would be obtained by using a discriminant function analysis on these data, as previously suggested by other studies.</p><p><b> Results:</b> A histological assessment of CD mucosa confirmed the morphological changes to the duodenum associated with the degree of CD severity. Discriminant function analysis of the histological data was then used to develop CD classification equations which accurately categorised CD patients from healthy control patients. However, these histology-based equations had low classification resolution and could not discriminate patients into individual Marsh score categories. To examine gene expression changes; a CD-specific qPCR array was constructed which showed a total of 25 genes to be significantly differentially expressed in CD patients. As expected, Th1 immune genes such as interferon gamma were strongly associated with CD severity. Definition of discriminant equations using duodenal gene expression data then demonstrated the reliable classification of CD patients into the different Marsh score categories. The utility of classification equations defined using empirically-derived histology and gene expression data was then further explored using <i>in silico</i> modelling of simulated data. The most accurate simulated prediction of CD severity was achieved using equations defined by both histological and gene expression data. The applicability of these classification equations to correctly categorise patients with an equivocal CD diagnosis was then shown using two case studies.</p><p><b> Conclusions:</b> Thus, this pilot study has defined discriminant equations which can objectively classify CD patients correctly into Marsh score categories with high resolution. Moreover, these equations proved useful in classifying patients with an equivocal CD diagnosis into a discrete Marsh score category. Finally the high resolution objective classification of CD patients into discrete Marsh score categories may be used to monitor disease progression and treatment effectiveness in CD patients. However the equations defined in this pilot study need to be confirmed in a much larger study which contains both CD and non-CD duodenal pathologies before a clinical diagnostic test can be defined.</p>