Modelling of the transcriptome using networks
| Main Authors: | Azevedo, Tiago, Dimitri, Giovanna Maria, Lio, Pietro, Gamazon, Eric R |
|---|---|
| Format: | info software |
| Terbitan: |
, 2020
|
| Online Access: |
https://zenodo.org/record/3842659 |
Daftar Isi:
- This repository contains all the code necessary to run and further extend the experiments presented in our paper. Please cite: Azevedo, Tiago, Dimitri, Giovanna Maria, Lio, Pietro, & Gamazon, Eric R. (2020) "Multilayer modelling and analysis of the human transcriptome." bioRxiv. https://www.biorxiv.org/content/10.1101/2020.05.21.109082v4 Azevedo, Tiago, Dimitri, Giovanna Maria, Lio, Pietro, & Gamazon, Eric R. (2020, May 25). Modelling of the transcriptome using networks (Version 1). Zenodo. http://doi.org/10.5281/zenodo.3842659 Abstract In the present work, we performed a comprehensive intra-tissue and inter-tissue network analysis of the human transcriptome. We generated an atlas of communities in co-expression networks in each of 49 tissues and evaluated their tissue specificity. UMAP embeddings of gene expression from the identified communities recovered biologically meaningful tissue clusters, based on tissue organ membership or known shared function. We developed an approach to quantify the conservation of global structure and estimate the sampling distribution of the distance between tissue clusters via bootstrapped manifolds. We found not only preserved local structure among clearly related tissues (e.g., the 13 brain regions) but also a strong correlation between the clustering of these related tissues relative to the remaining ones. Interestingly, brain tissues showed significantly higher variability in community size than non-brain (p = 1.55x10-4). We identified communities that capture some of our current knowledge about biological processes, but most are likely to encode novel and previously inaccessible functional information. For example, we found a 17-member community present across all of the brain regions, which shows significant enrichment for the nonsense-mediated decay pathway (adjusted p = 1.01x10-37). We also constructed multiplex architectures to gain insights into tissue-to-tissue mechanisms for regulation of communities in the transcriptome, including communities that are likely to play a functional role throughout the central nervous system (CNS) and communities that may participate in the interaction between the CNS and the enteric nervous system. Notably, new gene expression data can be embedded into our models to accelerate discoveries in high-dimensional molecular datasets. Our study provides a rich resource of co-expression networks, communities, multiplex architectures, and enriched pathways in a broad collection of tissues, to catalyse research into inter-tissue regulatory mechanisms and enable insights into their downstream phenotypic consequences.