Advancing clinical differentiation between constipation-predominant irritable bowel syndrome and functional constipation using machine learning models

Objectives

A growing volume of literature reports that there are signature associations between microbial abundance and functional bowel disorders. Given that we have secured a source of IBS-C/FC microbial sequencing and clinical data, we propose to develop a data-driven machine learning model that can discriminate between the two disorders.

Design and subjects

To meet the mimimum sample size according to our power analysis, we plan to obtain genotypic, enterotypic and clinical data for 70 FC individuals and 120 IBS-C individuals. We will then create, test and validate several machine learning algorithms using features extracted from the data we gathered. The expected outcome is to design a model that classifies IBS-C cases and FC cases more accurately than a random classifier would.

Significance

Even with modest accuracy, such a model would have immediate clinical relevance as a tool to support clinical differentiation between IBS-C and FC. Should our results be positive, it may inform the development of similar models to discriminate between other bowel disorders such as IBS-D and FD.