BRANEnet: embedding multilayer networks for omics data integration

Abstract

Background Gene expression is regulated at different molecular levels, including chromatin accessibility, transcription, RNA maturation, and transport. These regulatory mechanisms have strong connections with cellular metabolism. In order to study the cellular system and its functioning, omics data at each molecular level can be generated and efficiently integrated. Here, we propose BRANEnet, a novel multi-omics integration framework for multilayer heterogeneous networks. BRANEnet is an expressive, scalable, and versatile method to learn node embeddings, leveraging random walk information within a matrix factorization framework. Our goal is to efficiently integrate multi-omics data to study different regulatory aspects of multilayered processes that occur in organisms. We evaluate our framework using multi-omics data of Saccharomyces cerevisiae, a well-studied yeast model organism. Results We test BRANEnet on transcriptomics (RNA-seq) and targeted metabolomics (NMR) data for wild-type yeast strain during a heat-shock time course of 0, 20, and 120 min. Our framework learns features for differentially expressed bio-molecules showing heat stress response. We demonstrate the applicability of the learned features for targeted omics inference tasks: transcription factor (TF)-target prediction, integrated omics network (ION) inference, and module identification. The performance of BRANEnet is compared to existing network integration methods. Our model outperforms baseline methods by achieving high prediction scores for a variety of downstream tasks.

Network dynamics and therapeutic aspects of mRNA and protein markers with the recurrence sites of pancreatic cancer

“…Stable associations, including inter- and intra-connections between proteome and transcriptome data, across distinct PDAC recurrences were identified. While multi-layer associations were also identified using pancreatic cancer TCGA datasets by Jagtap et al , the associations revealed in this study did not overlap with those found in this study [ 32 ]. The observed difference in associations identified could be attributable to the use of mRNA and protein data sets in this study, compared to the use of miRNA and mRNA data sets to perform network-based multi-omics integration.…”

“…The observed difference in associations identified could be attributable to the use of mRNA and protein data sets in this study, compared to the use of miRNA and mRNA data sets to perform network-based multi-omics integration. Alternatively, the difference could be due to a Spearman partial correlation threshold being used in this study compared to the use of a Pearson correlation threshold and biological a priori knowledge for important association determination and network construction [ 32 ]. Consequently, the strength of monotonic associations between features were measured using Spearman correlation as compared with linear associations using Pearson correlation [ 53 ].…”

“…On the other hand, Jagtap et al, used multi-omics network-based methodologies to perform microRNA (miRNA) and mRNA inference using TCGA pancreatic data sets [ 31 ]. Additionally, biological prior knowledge was included in the BRANET model, the performance of BRANET miRNA-mRNA regulatory network inference was assessed [ 32 ]. On the other hand, Becker et al, developed Correlation Analysis of Large-scale (biological) Systems (CorALS), a Python package that supported correlation analysis, feature selection using highly correlated variable (feature) pairs, and network construction using selected features from high-dimensional multi-omics pregnancy data sets [ 33 ].…”

Contextual AI models for single-cell protein biology

“…These approaches can resolve protein interaction networks across tissues [29][30][31][32][33] and cell types by integrating molecular cell atlases [34,35] and extending our understanding of the relationship between protein and function [36][37][38]. Protein representation learning methods can predict multicellular functions across human tissues [31], design target-binding proteins [39][40][41][42] and novel protein interactions [43], and predict interactions between transcription factors and genes [37,38].…”

Gene Self-Expressive Networks as a Generalization-Aware Tool to Model Gene Regulatory Networks

“…These techniques aim at inferring GRNs by considering heterogeneous sources of data simultaneously [ 2 ]. Indeed, besides using gene expression data, these methods also rely on TF binding site patterns, or Chromatin Immuno-Precipitation data (e.g., [ 3 , 4 ]).…”