New Zealand Statistical Association 2024 Conference

Mario Prado

AgResearch

Navigating the maze of multi-omics methods for classification including microbiome data

Recent technological advancements have enabled the generation of various types of "omic" datasets, such as genomics, transcriptomics, proteomics, and metabolomics, which facilitates the discovery of new connections and functions in complex biological systems. Multi-omic analysis allows to study samples at a holistic level by integrating different data types. Here, we focus on multi-omics analysis that include microbiomes in the omics cascade, as they represent a fundamental part to the correct functioning of the ecosystem they inhabit.

The number of software methods available for multi-omics analysis has expanded exponentially with the growth of omic datasets, where these tools all have different assumptions, algorithms and are designed for specific applications. Consequently, selection of a suitable tool can be confusing for researchers, while methods for multi-omics analysis with microbiome data are lacking. We provide a review of the classification methods publicly available to integrate two or more omic layers (N-integration) for classification that supports microbiome data as input. As part of this review, we also evaluate a selection of these tools on real data sets – human gut, soil composition, coral holobiont, among others – to understand their performance across various factors (e.g., types of microbiomes, degree of group separation, sample size, number of omic layers). We found that classification tasks can benefit from a multi-omic analysis, their accuracy increased compared to a reductionist perspective. Additionally, we were able to compare the tools’ different approaches and how they react to microbiome data as new multi-omic input and its characteristics.