Abstract

The integration and analysis of multi-omics modalities is an important challenge in bioinformatics and data science in general. A standard approach is to conduct a series of univariate tests to determine the significance for each parameter, but this underestimates the connected nature of biological data and thus increases the number of false-negative errors. To mitigate this issue and to understand how different omics’ data domains are jointly affected, we used the Stacked Regularization model with Bayesian optimization over its full parameter space. We applied this approach to a multi-omics data set consisting of microbiota, metabolites and clinical data from two recent clinical studies aimed at detecting the impact of replacing part of the vegetable fat in infant formula with bovine milk fat on healthy term infants. We demonstrate how our model achieves a high discriminative performance, show the advantages of univariate testing and discuss the detected outcome in its biological context.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call