A Wavelet-Based Approach Reveals Host-Cell-Type-Specific Multi-omics Networks in Inflammatory Bowel Disease

Abstract

The interplay between microbiome metabolites and human cells is crucial and mechanistically linked to human health and disease. Inflammatory bowel disease (IBD) is linked to microbiome metabolite and host gene expression. However, details of the microbiome and host interplays remain elusive. We carry out microbiome metabolome-wide and host transcriptome-wide association studies of IBD with microbiome metabolite targeted cell types discovery via leveraging the publicly available IBD data sets from Human Microbiome Project 2 (HMP2). By performing deconvolution on the transcriptomic data and applying discrete wavelet transform (DWT), we obtained cell type-metabolite correlations, which we visualize in the form of heatmaps and networks. We also carried out both targeted and untargeted approaches by mean of correlating the microbiome data matrix to host transcriptomic data. Given the limited sample size, in addition to visualizing a global picture of the interplay landscape between microbiome metabolites and host genes alongside distinct clusters of IBD and healthy controls with UMAP and t-SNE, we found a set of microbiome metabolites most likely linked to IBD and the transcriptomic signature of IBD. For the targeted approach, we also refer to the single-cell gene signature dataset, i.e., the MSigDB C8, uncovering a bile acid, namely, lithocholate, targeted cell types, including intestine lymphoid cells and enterocytes. Moreover, we utilized Mendelian Randomization causality tests with ursodeoxycholic acid (UDCA), RUNX1 gene, and IBD, resulting in a putative causality network of RUNX1, UDCA, and IBD. Taken together, our approaches shed light on the mechanistic interplay of microbiome metabolites and host cells in human health and disease.