Prediction of the gut microbiome from plasma metabolites

Project description

The gut microbiome plays an important role in human health and has been associated with various conditions, including gastrointestinal and cardiometabolic diseases. However, many studies that have collected unique phenotypes or sampled from unique populations have historically not collected fecal samples. To address this, we aim to predict the gut microbiome using metabolites measured in plasma.

We are using deep shotgun metagenomics data and untargeted plasma metabolome profiles from 8,651 participants of the SCAPIS-Uppsala and SCAPIS-Malmö. Two types of prediction models will be employed: (a) a nonparametric model using gradient-boosted decision trees and (b) a linear model using ridge regression, both assessed through a ten-fold cross-validation process. For binary outcomes (presence or absence of species), we will evaluate predictive performance using the area under the curve (AUC). For continuous outcomes such as species abundance and alpha diversity, we will evaluate performance using R². We hypothesize that our results can enrich important studies without fecal samples with valuable microbiome information, such as an estimate of how diverse a participant’s microbiome is.