In some diseases such as Inflammatory Bowel Disease (IBD); if the disease is not diagnosed and monitored closely, the results can be very severe, such as the removal of the patient’s colon. On the other hand, currently existing most reliable diagnosis methods are very invasive (e.g. colonoscopy). An alternative approach for diagnosis can be abundance analysis of the microbial sample taken from the patients’ colon. This study aims to predict the disease state of the subject from bacterial abundances in stool samples taken from the patient.
105 samples were collected for this study among the patients of Dr. Athos Boudvaros; some of them
displaying IBD symptoms and others different diseases (control group). In Figure 6.3, each row block represents a set of bacterial groups at a taxonomic level (phylum level at the top and genus level at the bottom) and each column block represents a different patient group: control patients, Crohn’s disease (CD), and ulcerative colitis (UC). The only significant single biomarker was E. Coli, which is not seen in control and CD patients but seen in about a third of the UC patients. There seems to be no other single bacterial group that gives significant classification between the patient groups from these abundance measures.
Since E. Coli abundance is not a clear-cut single bacterial biomarker, using it as a diagnostic tool would yield low accuracy classification. On the other hand, we can take the entire bacterial group abundance distribution and feed them into a random forest and estimate cross-validation accuracy. After the classification method was employed, it was able to tell with 90% accuracy if the patient is diseased or not. This suggests that it is a competitive method with respect to other non-invasive diagnotic approaches which are generally highly specific but not sensitive enough.
One key difference between control and disease groups is the decrease in the diversity of the ecosystem. This suggests that the disease status is not controlled by a single germ but the overall robustness and the resilience of the ecosystem. When diversity in the ecosystem decreases, the patient might start showing disease symptoms.