Source-linked AI summary
Network Medicine in the age of biomedical big data
Abhijeet R. Sonawane, Scott T. Weiss, Kimberly Glass, Amitabh Sharma
TL;DR
Biomedical big data creates a need for network-based ways to understand phenotype-specific biology. This review surveys data sources and network types, organizes applications into three paradigms, and concludes that network medicine offers customized approaches for leveraging biomedical data.
Problem
The review examines how biological networks can be combined with large biomedical datasets to understand phenotypes amid expanding molecular-profile generation.
Method
The review surveys network types and biomedical data sources, then presents three paradigms for integrating phenotype-specific multi-omic information into PPI, coexpression, and gene regulatory networks.
Results
Network medicine provides customized ways to leverage biomedical data, with method selection dictated by the biological inquiry, hypotheses, study design, and available data.
Takeaways & Limitations
The review offers a resource for network scientists and biologists seeking to understand network modeling of biomedical data.
Takeaways & Limitations
Differential gene expression identifies disease-affected genes but does not reveal how those genes influence or are influenced by other genes.
Abstract
from arXiv · showhide
Network medicine is an emerging area of research dealing with molecular and genetic interactions, network biomarkers of disease, and therapeutic target discovery. Large-scale biomedical data generation offers a unique opportunity to assess the effect and impact of cellular heterogeneity and environmental perturbations on the observed phenotype. Marrying the two, network medicine with biomedical data provides a framework to build meaningful models and extract impactful results at a network level. In this review, we survey existing network types and biomedical data sources. More importantly, we delve into ways in which the network medicine approach, aided by phenotype-specific biomedical data, can be gainfully applied. We provide three paradigms, mainly dealing with three major biological network archetypes: protein-protein interaction, expression-based, and gene regulatory networks. For each of these paradigms, we discuss a broad overview of philosophies under which various network methods work. We also provide a few examples in each paradigm as a test case of its successful application. Finally, we delineate several opportunities and challenges in the field of network medicine. Taken together, the understanding gained from combining biomedical data with networks can be useful for characterizing disease etiologies and identifying therapeutic targets, which, in turn, will lead to better preventive medicine with translational impacts on personalized healthcare.
1 Introduction
Network medicine combines graph-based representations of biological components with large-scale biomedical data to explain phenotypes, cellular organization, and disease etiology. Advances in high-throughput technologies enable multi-omics datasets and motivate three paradigms for integrating biological networks with biomedical big data.
- 1 Introduction: Accurate measurements of molecular abundance profiles are needed to understand phenotypes, cellular function, tissue function, disease etiology, and cellular organization.Biomedical data analysis can help explain important features of molecular interactions.
- 1 Introduction: Network biology uses graph theory, systems biology, and statistical analyses to study holistic relationships among biological components.Its quantitative tools can characterize cellular organization and capture perturbation effects on intracellular networks.
- 1 Introduction: High-throughput technologies and declining sequencing costs have enabled massive multi-omics biomedical datasets characterizing diverse phenotypes.Examples include exome and whole-genome sequencing, transcriptomics, and proteomics.
- 1 Introduction: The paper presents three paradigms for combining biological networks with biomedical big data to understand phenotypes.These paradigms provide the framework for the review’s subsequent discussion.
2 Biomedical data sources
Advances in sequencing and large biomedical resources have enabled massive molecular-profile datasets spanning diverse phenotypes and diseases. These data support deeper biological-system analysis and motivate versatile network-medicine methods tailored to specific biological and disease contexts.
- Biomedical data sources: Sequencing advances and reduced per-base-pair costs have enabled massive molecular-profile data generation across diverse phenotypes and diseases.The HapMap project added an extensive catalogue of common human genetic variants based on microarray data after the Human Genome Project.
- Biomedical data sources: The Human Protein Atlas, Human Cell Atlas, and UK Biobank provide complementary resources spanning protein expression, single-cell omics, and human biomedical data.The Human Protein Atlas covers cells, tissues, pathologies, and 17 cancer types, while the Human Cell Atlas aims to map human cells and cell types.
- Biomedical data sources: Biomedical data enable deeper probing of biological systems and inspire methods that maximize information extraction while adapting to biological or disease context.Network-medicine tools are described as highly versatile and customizable, although collecting large-scale multi-time-point multi-omics data across disease conditions is expensive.
3 Primer on biological networks
Biological network analysis identifies system entities as nodes and their interactions as edges using multiple data sources. It also characterizes networks through their local, global, and mesoscale topological properties across diverse biological network types.
- 3 Primer on biological networks: Network-based studies define critical biological entities as nodes and the interactions between them as edges, often integrating multiple data sources.Protein–protein interaction networks, or interactomes, represent proteins and their physical interactions.
- 3 Primer on biological networks: Biological networks encompass protein-interaction, metabolic, ecological, microbiome, genotype, and transcriptomic representations.Metabolic networks capture biochemical interactions between metabolites and enzymes, while ecological networks can represent microbial interactions in microbiome data.
- 3 Primer on biological networks: Network analysis evaluates local and global topology by identifying modulators, driver nodes, network structures, and properties such as degree distribution, path length, clustering, diameter, and controllability.These measures support characterization and comparison of network topologies.
- 3 Primer on biological networks: Mesoscale analysis uses subgraphs and network motifs—recurrent patterns connecting typically three or four nodes—and extends them to larger graphlets for interactome analysis.Enriched connectivity patterns can be identified by assessing whether structures are over-represented in a network.
4 Integrating data with networks/Combining biomedical data with networks:
This section frames network medicine as a way to integrate phenotype-specific biomedical data with biological networks to generate lab-testable hypotheses. It organizes the review around PPI, GCN, and GRN paradigms, emphasizing nuanced, condition-specific analysis.
- Combining biomedical data with networks:: Integrating biomedical data with networks aims to identify biomolecular entities and interaction changes that explain observed phenotypes and generate lab-testable hypotheses.The approach requires relevant data and appropriate network analysis in tissue-, cell-, or disease-specific environments.
- Combining biomedical data with networks:: PPI analysis combines baseline protein interactions with disease information to identify critical modules formed when defects or mutations propagate through the network.Phenotype-specific interactions are added as an extra layer from separate biomedical data.
- Combining biomedical data with networks:: GCNs are inherently context-specific because they are constructed from correlations in a given gene-expression dataset, whereas GRNs often begin with sequence-derived potential regulatory interactions.GRN baseline edges can be inferred from transcription-factor binding motifs in gene promoters.
- Combining biomedical data with networks:: The review examines three network types—PPI networks, GCNs, and GRNs—and presents exemplar methods and complementary philosophies for each.These paradigms frame how phenotype-specific molecular information is embedded into network models.
- Combining biomedical data with networks:: Applying network phenomenology to biomedical big data requires a nuanced, condition-specific approach, with each paradigm evaluated through its questions, examples, and outcome diagnostics.The review discusses these paradigms separately to explain how they embed phenotype-specific molecular information into networks.
interactome
High-throughput interactome maps support disease-gene prioritization and identification of disease-related network components, while phenotype-specific expression networks help reveal disease-associated rewiring. Integrating increasingly comprehensive interaction, genetic, phenotypic, and multi-omic data remains important for improving network-based disease analysis.
- Data integration: Combining yeast-2-hybrid, AP-MS, crosslinking AP-MS, curated interaction resources, and large genetic-phenotypic cohorts can address missing interactions and disease genes.Multi-omic integration into network-inference models remains an open challenge, while transcript-based networks leverage biomedical big data to derive disease-specific information.
- Disease-gene prediction: Interactome topology supports prioritizing novel disease-associated genes and proteins by relating them to known disease candidates through molecular interaction hierarchies.The PPI network serves as a map of potential biological interactions for identifying genes involved in tissue regulation or disease dysregulation.
- Disease modules: Disease-related subnetworks provide a broader space for discovering pathways and mechanisms, with connectivity significance among seed-associated proteins proposed as a predictive quantity.The approach identifies close neighbors of known disease-associated proteins using interactome topology.
- Phenotype-specific networks: Phenotype-specific networks enable condition-to-condition comparisons that can uncover pathway rewiring induced by disease, treatment, or environmental stimuli.Constructing separate networks for each condition can mitigate systematic experimental biases and errors.
5 Conclusion and future directions
The review presents network medicine as a framework for integrating biomedical big data, with method selection guided by the biological inquiry, hypotheses, study design, and available data. It highlights network-based multi-omics integration as a promising approach for explaining disease etiologies and cellular function.
- 5 Conclusion and future directions: Network medicine provides customized ways to leverage biomedical data, with appropriate method choice dictated by the biological inquiry, hypotheses, study design, and available data.
- 5 Conclusion and future directions: The framework can integrate heterogeneous omics data by elucidating mutual influences among molecular data types to help explain disease etiologies and cellular function.
- 5 Conclusion and future directions: Multi-omics data integration using networks has gained substantial scientific attention, alongside newer tools including multiplex networks, network fusion, innovative community detection, and higher-order structural modularity.
7 Author Contributions
All listed authors substantially contributed to the paper’s intellectual development, writing, and editing, and approved the manuscript for publication.
- 7 Author Contributions: All listed authors contributed substantially to the intellectual work, writing and editing, and approved the manuscript for publication.
8 Funding
The authors acknowledge support from NIH/NHLBI and other NIH grants, while stating that the funders had no role in the research or publication process.
- 8 Funding: K. Glass was supported by NIH/NHLBI grant K25HL133599, and the study received NIH grants R01 HL118455-04-1 and P01 HL13285.The funders had no role in study design, data collection and analysis, the decision to publish, or manuscript preparation.