Computational Molecular Biology 2024, Vol.14, No.2, 64-75 http://bioscipublisher.com/index.php/cmb 72 Table 1 Description of different tools for multi-omics integration with their application and their major strength and limits (Adopted from Raufaste-Cazavieille et al., 2022) Method Principle Aim Omics element Pros Cons JIVE Matrix factorization Disease subtyping systemic knowledge, module detection Genomics and epigenomics Integrate large amount of data Sensitive tooutliers and missing values NMF - Disease subtyping module detection, biomarker discovery Genomics and epigenomics Filtering weak signal.Integrate large amount of data. Detection of cluster of small size Time and memory consuming. Underperforming on missing values nNMF - - - - - jNMF - - - - - intNMF - - - - - SLIDE - Disease subtyping, module detection, biomarker discovery Genomics,epigenom ics and proteomics Integrate large amount of data Underperforming with missing values. Optimum solution is not guaranteed MALA Logicdata mining Sample classification Genomics and transcriptomics Works well on experimental data Integrate large amount of data Phenotype number must be delivered with data. Sensitive to missing values iCluster Gaussian latent variable model - Genomics,epigenom ics and transcriptomics - Needs to test a large amount of solution tofind the most relevant iCluster+ Generalized linear regression Disease subtyping Genomics,transcript omics, proteomics and epigenomics Handle missing values No evaluation of statistical significancefor selected features iClusterBayes Bayesian integrative clustering Biomarker discovery Genomics,transcript omics, and epigenomics Good performance in thepresence of explicative data Underperform with outliers MOFA Bayesian factor analysis Biomarker discovery, systemic knowledge Proteomics,metabolo mics and lipidomics Handle well missing values Linear model can miss linear relation MOFA+ - - Genomics and epigenomics The use of continuous learning enabling MOFA to recover different trajectory Need of multi-moda measurement for the same set of cells Table caption: JIVE:joint and individual variation explained; (n,j,int) NMF:(network,joint,integrative) non-negative matrix factoization; SLIDE: structural learning and integrative decomposition; MALA: micro array logic analyzer; MOFA:multi-omics factor analysis (Adopted from Raufaste-Cazavieille et al., 2022) 7.3 Microbiome and host interactions The study of microbiome and host interactions has greatly benefited from multi-omics integration. By combining metagenomics, metatranscriptomics, metaproteomics, and metabolomics data, researchers can gain a comprehensive view of microbial communities and their functions. This approach has been particularly useful in understanding the role of the microbiome in health and disease. One successful example is the use of dynamic Bayesian networks (DBNs) to integrate multi-omics data from longitudinal microbiome studies. This method has been applied to data collected from patients with inflammatory bowel disease (IBD), allowing researchers to identify known and novel interactions between microbial taxa, their genes, metabolites, and host genes. The resulting models have provided insights into the temporal interactions and their impact on host expression, which are crucial for understanding disease progression and developing targeted
RkJQdWJsaXNoZXIy MjQ4ODYzNA==