Falsifiable Network Models. A Network-based Approach to Predict Treatment Efficacy in Ulcerative Colitis

Abstract

This work is focused on understanding the treatment efficacy of patients with ulcerative colitis (UC) using a network-based approach. UC is one of two forms of inflammatory bowel disease (IBD) along with Crohn’s disease. UC is a debilitating condition characterized by chronic inflammation and ulceration of the colon and rectum. UC symptoms occur gradually rather than abruptly, and the degree of symptoms differs across UC patients. Only around 20% of all UC cases can be explained by known genetic variations, implying a more ambiguous aetiology that is yet not fully understood but is thought to involve a complex interplay between genetic and environmental factors. The available therapy for UC substantially reduces symptoms and achieves long-term remission. However, about one-third of UC patients fail to respond to anti-TNFα therapy and consequently develop long-term side effects due to medication. Non-response to existing antibody-based therapies in subgroups of UC patients is a major challenge and incurs a healthcare burden. Therefore, the disease markers for predicting therapy response to assist individualized therapy decisions are needed. To date, no quantitative computational framework is available to predict treatment response in UC. We developed a quantitative framework that uses gene expression data and existing biological background information on signalling pathways to quantify network connectivity from receptors to transcription factors (TF) that are involved in UC pathogenesis. Variations in network connectivity in UC patients can be used to identify responders and non-responders to anti-TNFα and anti-Integrin treatment. Our findings allow us to summarize the effect of small gene expression changes on the overall connectivity of a signalling network and estimate the effect this will have on the individual patients' responses. Estimating the network connectivity associated with varied drug responses may provide an understanding of individualized treatment outcomes. Our model could be used to generate testable hypotheses about how individual genes act together in networks to cause inflammation in UC as well as other immune-inflammatory diseases such as psoriasis, asthma, and rheumatoid arthritis.