The multi-omics is biological analysis approach to combine different omics data such as genomics, proteomics, and metabolomic. It provides an integrated perspective to better understand complicated biological events by observing and interpreting a large amount of information focused on the association between different parameters.

In this study, we constructed various algorithms and statistical analysis that can link metabolome and microbiome to prioritizing microbial species.

LC-MS Orbitrap based untargeted metabolic analysis and Shotgun metagenomic sequencing were performed in Healthy controls and non-alcoholic fatty liver disease (NAFLD) patients fecal samples.

For metabolomic data, we built retention time prediction model using five deep & machine learning algorithms in R and python. The prediction models allowed the accurate prediction of compounds from the unknown peak.

In addition, we optimized the computational process for the association between metabolome and microbiome data based on SPARCC algorithm, which guaranteed robust and reliable correlation coefficient among heterogenous data matrix.

Finally, we built cross-validation process based on leave-one-out algorithm in R. The system enhanced validation step for the contribution of prioritized candidate (microbial species) to KEGG modularity, metabolome, and host phenotype.