Keywords: microbiome, high dimension, semi-parametric models, between-subject attributes
1. Background
The human microbiome plays an important role in human disease and health. Identifying those factors that are associated with microbiome composition not only provides insights on the inherent disease mechanism, but also gives suggestions on modulating the microbiome composition for therapeutical purposes.
Chronic alcohol consumption causes an increased intestinal permeability and changes in the intestinal microbiota composition which contributes to the development and progression of alcoholic liver disease (ALD). We investigated the intestinal microbiota in a well described cohort of alcoholic hepatitis patients (AH), alcohol use disorder (AUD) patients and healthy controls (HC), in order to find whether diagnostic group is significantly associated with microbiome composition.
2. Methods
In order to assess human microbiome composition, we first extract a community DNA sample and then amplify and sequence the 16S rRNA gene. The highly similar sequences are binned into Operational Taxonomic Units (OTUs) to obtain OTU abundances of each sample. However, testing the association of microbiome composition with potential phenotypical outcomes using OTU abundances is difficult due to high dimensionality and non-normality of the OTU data.
Instead, distance-based measures are used to compare pairs of microbiome samples to define diversity in microbiome composition between individuals, or beta-diversity. Analysis of such “between-subject” attributes are not amenable to the predominant “within-subject” based statistical paradigm, such as Pearson’s correlation and linear regression. We developed a new approach to perform valid inference by utilizing the functional response models (FRM), a class of semi-parametric models for between-subject as well as within-subject attributes.
3. Results & Conclusions
By modeling beta-diversity using the new FRM-based approach, we found that diagnostic group (AH/AUD/HC) is a significant factor associated with microbiome