Signaling Networks and Disease Modules

Summary

Covers cell signaling cascades as directed graphs, network medicine concepts including disease modules and network proximity, drug target discovery, drug repurposing via knowledge graphs, cancer genomics, and precision medicine applications.

Concepts Covered

This chapter covers the following 30 concepts from the learning graph:

Cell Signaling Cascade
Signal Transduction
Receptor
Kinase Cascade
Second Messenger
Directed Signaling Graph
Feedback Loop
Feed-Forward Loop
Network Medicine
Disease Module
Network Proximity
Guilt by Association
Drug Target
Drug Target Validation
Drug Repurposing
Drug-Target-Disease Graph
Graph Model for Repurposing
Pharmacogenomics
Cancer Driver Genes
Tumor Suppressor Gene
Oncogene
Cancer Network Analysis
Precision Medicine
Biomarker Discovery
Clinical Network Analysis
Side Effect Prediction
Drug-Drug Interaction Graph
Adverse Event Network
Comorbidity Network
Disease Gene Prioritization

Prerequisites

This chapter builds on concepts from:

Welcome, Explorers! Let's Connect the Dots!

Cells don't work in isolation -- they talk to each other and to themselves through intricate signaling cascades. When we draw those cascades as directed graphs, we unlock a powerful framework for understanding disease at the network level. In this chapter we will model cell signaling as a graph, discover disease modules hidden in the interactome, use knowledge graphs to repurpose existing drugs for new diseases, and apply cancer network analysis to identify the driver mutations that matter most. Let's connect the dots!

Cell Signaling: From Molecules to Directed Graphs

A cell signaling cascade is a chain of molecular events that transmits information from the cell surface to the nucleus or other intracellular targets. The process begins when an extracellular signal -- a hormone, growth factor, or cytokine -- binds to a receptor protein embedded in the plasma membrane. That binding event triggers a conformational change that activates intracellular proteins, often through phosphorylation, ultimately altering gene expression, metabolism, or cell behavior. The entire process of converting an external signal into an internal cellular response is called signal transduction.

Receptors are the gatekeepers of signaling. Receptor tyrosine kinases (RTKs) such as EGFR and VEGFR dimerize upon ligand binding and autophosphorylate their intracellular domains, creating docking sites for downstream adaptor proteins. G-protein-coupled receptors (GPCRs) activate heterotrimeric G proteins that release second messengers -- small diffusible molecules such as cyclic AMP (cAMP), calcium ions (Ca\(^{2+}\)), and inositol trisphosphate (IP\(_3\)) -- which amplify the signal and distribute it throughout the cell. Second messengers are critical because they allow a single receptor activation event to influence hundreds of downstream targets simultaneously.

Diagram: Cell Signaling Cascade Overview

View Signaling Cascade Overview MicroSim Fullscreen

Diagram Details

sim-id: signaling-cascade-overview Library: vis-network Status: Specified A directed graph showing a receptor at the membrane, adaptor proteins, a kinase cascade (RAF-MEK-ERK), second messenger nodes, and transcription factor targets in the nucleus. Edge arrows indicate activation. Feedback loops are drawn as dashed edges returning from downstream to upstream nodes.

Once a receptor is activated, the signal typically passes through a kinase cascade -- a series of protein kinases in which each kinase phosphorylates and activates the next. The RAS-RAF-MEK-ERK pathway is the canonical example: activated RAS recruits RAF to the membrane, RAF phosphorylates MEK, and MEK phosphorylates ERK. ERK then translocates to the nucleus and activates transcription factors that drive cell proliferation. Kinase cascades provide signal amplification (one active kinase molecule can phosphorylate many copies of the next kinase) and offer multiple points where the cell can regulate signal strength.

Directed Signaling Graphs and Network Motifs

The natural graph representation of a signaling pathway is a directed signaling graph \(G = (V, E)\) where each node \(v \in V\) represents a signaling molecule (receptor, kinase, transcription factor, second messenger) and each directed edge \((u, v) \in E\) represents a regulatory interaction -- activation or inhibition. Edge attributes encode the sign of the interaction (positive for activation, negative for inhibition) and may carry confidence scores derived from experimental evidence.

Two network motifs appear repeatedly in signaling graphs and have profound consequences for cellular behavior. A feedback loop is a directed cycle in the signaling graph where a downstream node influences an upstream node. Negative feedback loops, such as the phosphatase SHP-1 dephosphorylating an activated receptor, dampen the signal and restore the system to its resting state. Positive feedback loops, such as ERK phosphorylating and stabilizing SOS (an activator of RAS), create bistable switches that lock the cell into an all-or-nothing decision.

A feed-forward loop is a three-node motif in which node \(A\) regulates node \(B\), node \(B\) regulates node \(C\), and node \(A\) also directly regulates node \(C\). The coherent type 1 feed-forward loop (where all three edges are activating) acts as a persistence detector: node \(C\) is only activated if node \(A\) provides a sustained signal, because the direct path from \(A\) to \(C\) and the indirect path through \(B\) must both deliver their inputs before \(C\) responds. This motif filters out transient noise in the signaling environment and is enriched in transcriptional and signaling networks alike.

Motif	Structure	Biological Function	Example
Negative feedback loop	\(A \to B \to C \dashv A\)	Signal attenuation, homeostasis	SHP-1 dephosphorylating activated receptor
Positive feedback loop	\(A \to B \to C \to A\)	Bistable switching, memory	ERK stabilizing SOS in RAS pathway
Coherent feed-forward loop	\(A \to B \to C\), \(A \to C\)	Persistence detection, noise filtering	Growth factor signaling to cell cycle entry
Incoherent feed-forward loop	\(A \to B \dashv C\), \(A \to C\)	Pulse generation, adaptation	NFkB signaling with IkB inhibition

Think About It: Why Do Signaling Networks Need Feedback?

Without negative feedback, a single growth factor molecule could activate its receptor forever, driving continuous cell division. Without positive feedback, cells could not commit to irreversible decisions like apoptosis. The balance between these motifs determines whether a cell responds proportionally, switches sharply, or oscillates.

Network Medicine and Disease Modules

Network medicine is a framework that uses the human interactome -- the complete network of protein-protein interactions -- to understand and treat disease. Its foundational principle is that disease genes are not scattered randomly across the interactome. Instead, genes associated with a particular disease tend to cluster into a connected subgraph called a disease module. This clustering reflects the biological reality that diseases arise from the dysfunction of coherent molecular pathways rather than isolated gene failures.

The disease module hypothesis was formalized by Barabasi and colleagues, who demonstrated that disease-associated proteins identified through genome-wide association studies (GWAS) tend to be closer to each other in the interactome than expected by chance. To quantify this clustering, we measure the network proximity between two sets of nodes. Given a set of disease genes \(A\) and a set of drug target genes \(B\) in the interactome, the closest network proximity is defined as:

\[d_c(A, B) = \frac{1}{\|A\| + \|B\|} \left( \sum_{a \in A} \min_{b \in B} d(a, b) + \sum_{b \in B} \min_{a \in A} d(b, a) \right)\]

where \(d(a, b)\) is the shortest path distance between nodes \(a\) and \(b\) in the interactome. To assess statistical significance, we compare the observed proximity to a reference distribution generated by randomly selecting gene sets of the same size and degree distribution, computing a \(z\)-score:

\[z_c = \frac{d_c(A, B) - \mu_{\text{random}}}{\sigma_{\text{random}}}\]

A significantly negative \(z\)-score indicates that the two gene sets are closer in the network than expected by chance, suggesting a biological relationship between the disease and the drug targets.

Diagram: Disease Module in the Human Interactome

View Disease Module in the Human Interactome MicroSim Fullscreen

Diagram Details

sim-id: disease-module-interactome Library: vis-network Status: Specified An interactive network visualization of a subset of the human interactome. Disease-associated genes for two diseases (e.g., asthma and diabetes) are highlighted in distinct colors. The user can toggle disease modules on and off to see how they cluster within the larger network. Overlapping modules (shared genes between diseases) are highlighted, illustrating comorbidity.

The guilt-by-association principle extends the disease module concept to predict novel disease genes. If a protein of unknown function interacts with several known disease proteins, it is likely involved in the same disease pathway. The simplest scoring function assigns each candidate gene \(g\) a guilt-by-association score based on its neighbors:

\[S(g) = \frac{\sum_{n \in N(g)} w(n) \cdot \mathbb{1}[n \in D]}{|N(g)|}\]

where \(N(g)\) is the set of neighbors of gene \(g\) in the interactome, \(D\) is the set of known disease genes, \(w(n)\) is an optional weight reflecting confidence in the disease association, and \(\mathbb{1}[n \in D]\) is the indicator function. More sophisticated approaches use random walks with restart, network propagation, or graph neural networks, but the core logic remains the same: network context predicts gene function.

This guilt-by-association framework directly supports disease gene prioritization -- the task of ranking candidate genes from a GWAS locus by their likelihood of being true causal genes. Because GWAS identifies broad genomic regions that may contain dozens of genes, network-based prioritization narrows the list to those candidates that are well connected to known disease biology. Tools such as PRINCE, GeneMANIA, and DIAMOnD implement variations of this approach and are routinely used to prioritize genes for experimental follow-up.

Drug Targets and Drug Target Validation

A drug target is a molecular entity -- typically a protein -- whose activity can be modulated by a drug to produce a therapeutic effect. Identifying and validating drug targets is the first step in the drug discovery pipeline and one of its most failure-prone stages: approximately 90% of drug candidates that enter clinical trials ultimately fail, and a leading cause is insufficient target validation.

Drug target validation is the process of gathering evidence that modulating a target will have the desired therapeutic effect without unacceptable toxicity. Network medicine provides a powerful lens for validation. If a candidate target sits within the disease module of the condition being treated, it is more likely to influence disease biology. If it also has low network proximity to the disease modules of unrelated conditions, it is less likely to cause off-target effects. Network-based validation examines:

Module membership: Is the target within or adjacent to the disease module?
Centrality: How central is the target in the interactome? Highly central targets may affect too many pathways.
Pathway context: Does the target participate in the signaling cascade that is dysregulated in the disease?
Genetic evidence: Do loss-of-function variants in the target gene associate with disease phenotypes?

Think About It: The Centrality Trade-Off

A highly connected hub protein in the interactome might seem like an ideal drug target because it influences many processes. But that same connectivity means inhibiting it could disrupt essential cellular functions and cause severe side effects. The best drug targets are often proteins with moderate centrality that sit specifically within the disease module -- influential enough to matter, but not so connected that modulating them causes widespread collateral damage.

Drug Repurposing via Knowledge Graphs

Drug repurposing (also called drug repositioning) seeks new therapeutic uses for existing approved drugs. Because repurposed drugs have already passed safety testing, they can reach patients faster and at lower cost than novel compounds. The graph-based approach to repurposing has emerged as one of the most powerful strategies in computational drug discovery.

The central data structure is a drug-target-disease graph -- a heterogeneous knowledge graph with three primary node types (drugs, protein targets, and diseases) and multiple edge types connecting them. Edges encode relationships such as "drug binds target," "target associated with disease," "drug indicated for disease," "drug causes side effect," and "disease comorbid with disease." This tripartite structure is the key graph model for repurposing: if a drug treats disease \(A\), and disease \(A\) shares molecular mechanisms with disease \(B\), then the drug may also treat disease \(B\).

Diagram: Drug-Target-Disease Knowledge Graph

View Drug-Target-Disease Knowledge Graph MicroSim Fullscreen

Diagram Details

sim-id: drug-target-disease-graph Library: vis-network Status: Specified A heterogeneous knowledge graph with three node types: drugs (blue circles), protein targets (green diamonds), and diseases (red squares). Multiple edge types are shown in different colors: "binds" edges from drugs to targets, "associated with" edges from targets to diseases, "treats" edges from drugs to diseases, and "comorbid with" edges between diseases. The user can click on a drug node to highlight all its connections and explore potential repurposing paths.

Case Study 2: Drug Repurposing with Hetionet

Hetionet is a publicly available knowledge graph that integrates data from 29 biomedical databases into a single heterogeneous network containing over 47,000 nodes of 11 types and over 2.25 million edges of 24 types. Node types include genes, compounds (drugs), diseases, anatomical structures, biological processes, and more. This rich structure makes Hetionet an ideal platform for systematic drug repurposing.

The repurposing algorithm works by extracting features from the graph structure. For each drug-disease pair, the algorithm computes path-based features: how many metapaths of each type connect the drug to the disease. A metapath is a sequence of edge types that defines a semantic route through the knowledge graph. For example, the metapath Compound-binds-Gene-associates-Disease (CbGaD) captures the traditional pharmacology route: a drug binds a protein that is genetically associated with a disease. The metapath Compound-treats-Disease-resembles-Disease (CtDrD) captures the repurposing-by-similarity route: a drug that treats one disease may also treat a similar disease.

Himmelstein and colleagues trained a logistic regression model on these metapath-based features using known drug-disease treatment pairs as positive examples. The model learned which graph connectivity patterns are predictive of therapeutic efficacy and then scored all unindicated drug-disease pairs. The approach recovered known treatments with high accuracy and generated novel predictions, several of which have since been supported by independent evidence. The key insight from Case Study 2 is that the structure of the knowledge graph itself -- the pattern of connections between drugs, targets, and diseases -- carries predictive information about therapeutic potential.

Tip: Explore Hetionet Yourself

Hetionet is freely available at het.io and can be loaded into Neo4j for Cypher queries. Try querying for all metapaths of length 3 or fewer between a drug you know (e.g., metformin) and a disease of interest. The number and diversity of connecting paths can reveal surprising biological connections.

Pharmacogenomics extends the drug-disease graph by incorporating genetic variation. Some patients metabolize a drug rapidly because they carry a fast variant of a cytochrome P450 enzyme, while others metabolize it slowly and experience toxicity at standard doses. By adding pharmacogenomic edges (gene variant -- alters response to -- drug) to the knowledge graph, we can predict which patient subgroups will benefit most from a given therapy and which will be at risk for adverse events.

Cancer Genomics: Drivers, Suppressors, and Oncogenes

Cancer is fundamentally a disease of the genome, but understanding which mutations actually drive tumor growth requires network analysis. The genome of a typical tumor harbors thousands of somatic mutations, but only a handful are cancer driver genes -- genes whose mutations confer a selective growth advantage to the tumor cell. The remaining mutations are passengers that accumulated by chance during cell division and have no functional consequence.

Cancer driver genes fall into two functional classes. A tumor suppressor gene encodes a protein that normally restrains cell growth, promotes DNA repair, or triggers apoptosis. When both copies of a tumor suppressor are inactivated by mutation (the "two-hit" hypothesis), the cell loses an essential brake on proliferation. TP53, RB1, and BRCA1 are well-known tumor suppressors. In contrast, an oncogene is a gene that, when activated by a gain-of-function mutation, drives cell proliferation. Oncogenes typically require mutation of only one copy (dominant). KRAS, BRAF, and MYC are classic oncogenes.

Case Study 4: Cancer Driver Genes via PPI Networks

Distinguishing drivers from passengers is the central challenge addressed in Case Study 4. The approach leverages the protein-protein interaction (PPI) network: if a mutated gene's protein product sits at a critical position in the interactome, disrupting it is more likely to have functional consequences than disrupting a peripheral protein.

Cancer network analysis integrates somatic mutation data from tumor sequencing with the PPI network. One widely used method scores each gene by combining its mutation frequency across patient cohorts with its network centrality and its proximity to known cancer genes. The algorithm proceeds as follows:

Map somatic mutations from a patient cohort onto the PPI network.
Score each gene by its mutation recurrence (how often it is mutated across patients).
Apply network propagation: spread the mutation signal through the network using a random walk with restart, so that genes near frequently mutated genes also receive elevated scores.
Rank genes by their propagated score and select those above a significance threshold as candidate drivers.

This approach identified novel driver genes missed by frequency-based methods alone. For example, genes that are rarely mutated but are direct network neighbors of multiple frequently mutated genes receive high propagated scores -- a pattern consistent with being essential pathway members whose loss can be partially compensated by other mutations. Case Study 4 demonstrates that PPI network topology is an essential complement to mutation frequency for driver gene identification.

Method	Input Data	Network Used	Strengths	Limitations
MutSigCV	Mutation frequency, covariates	None	Well-established statistics	Misses rare drivers
HotNet2	Mutations + PPI	PPI network	Finds subnetwork modules	Depends on PPI completeness
Oncodrive-FM	Functional impact of mutations	None	Considers mutation position	Ignores network context
Network propagation	Mutations + PPI	PPI network	Detects rare drivers near hubs	Sensitive to network quality

Precision Medicine and Clinical Network Analysis

Precision medicine aims to tailor treatment to the individual patient's molecular profile rather than applying one-size-fits-all therapies. The network medicine framework provides the computational infrastructure for this vision. By mapping a patient's genomic variants, gene expression profile, and clinical features onto the interactome, clinicians can identify which disease modules are activated and select therapies that target those specific modules.

Biomarker discovery is a key application of clinical network analysis. A biomarker is a measurable indicator of a biological state, such as a protein whose blood levels predict disease progression or a gene expression signature that classifies tumor subtypes. Network-based biomarker discovery identifies genes or proteins that occupy central positions within the disease module and whose expression levels correlate with clinical outcomes. These network biomarkers are often more robust than single-gene markers because they capture pathway-level activity rather than the noisy expression of individual genes.

Clinical network analysis integrates patient-level data (genomics, transcriptomics, proteomics, clinical records) with reference networks (interactome, signaling pathways, metabolic networks) to generate personalized disease models. For example, a patient's tumor transcriptome can be overlaid on a signaling network to identify which pathways are aberrantly activated, guiding selection of targeted therapies that inhibit those specific pathways.

The core insight of precision medicine is that the same clinical diagnosis -- say, non-small-cell lung cancer -- may involve different driver mutations, different activated signaling pathways, and different drug sensitivities across patients. Network analysis reveals these patient-specific differences and points toward individualized treatment strategies rather than one-size-fits-all protocols.

Side Effects, Drug Interactions, and Adverse Events

The same network framework that predicts therapeutic efficacy can also predict unintended consequences. Side effect prediction uses the interactome to identify off-target effects: if a drug's primary target sits near proteins involved in unrelated physiological processes, modulating the target may perturb those processes as well. By computing the network proximity between a drug's target set and the gene sets associated with various phenotypes, we can predict which side effects a drug is likely to cause before clinical trials begin.

A drug-drug interaction graph models the combinatorial space of polypharmacy. Nodes represent drugs and edges connect drugs that interact -- either pharmacokinetically (one drug alters the metabolism of another) or pharmacodynamically (two drugs modulate the same pathway in conflicting ways). Patients on multiple medications face exponentially growing interaction risks, and graph analysis can flag dangerous combinations by identifying cliques, dense subgraphs, or short paths in the interaction network.

Diagram: Drug-Drug Interaction Network

View Drug-Drug Interaction Network MicroSim Fullscreen

Diagram Details

sim-id: drug-drug-interaction-network Library: vis-network Status: Specified An interactive network showing drug nodes colored by therapeutic class. Edges represent known drug-drug interactions, with edge color indicating severity (red for severe, orange for moderate, yellow for mild). The user can search for a drug and see all its interactions highlighted. Dense clusters indicate drug classes with many internal interactions.

An adverse event network aggregates post-market surveillance data (such as FDA Adverse Event Reporting System data) into a bipartite graph linking drugs to reported adverse events. Projecting this network onto the drug dimension reveals which drugs share unusual side effect profiles -- a signal that they may share off-target mechanisms. Projecting onto the adverse event dimension reveals which side effects co-occur, potentially reflecting shared underlying biology.

Comorbidity Networks and Disease Relationships

A comorbidity network connects diseases that co-occur in patients more often than expected by chance. Edges are typically weighted by the relative risk or phi-correlation between two diseases measured across large patient populations. The structure of the comorbidity network reveals non-obvious relationships between diseases. For example, the observation that diabetes patients have elevated risk for Alzheimer's disease led researchers to investigate shared molecular mechanisms -- and indeed, insulin signaling pathways are disrupted in both conditions.

Comorbidity networks bridge the gap between clinical observation and molecular mechanism. When two diseases are comorbid and their disease modules overlap in the interactome, the shared genes suggest a mechanistic explanation for the comorbidity. When the modules are close but non-overlapping, the shared pathway neighborhood suggests that perturbation of one module propagates through the network to affect the other.

Network Type	Nodes	Edges	Primary Application
Drug-target-disease graph	Drugs, proteins, diseases	Binds, associates, treats	Drug repurposing
Drug-drug interaction graph	Drugs	Known interactions	Polypharmacy safety
Adverse event network	Drugs, adverse events	Reports	Side effect prediction
Comorbidity network	Diseases	Co-occurrence statistics	Disease relationship discovery
Disease module	Proteins within interactome	PPI edges	Mechanistic disease understanding

Watch Out: Correlation Is Not Causation in Comorbidity Networks

Comorbidity networks are built from observational data, and shared risk factors (age, smoking, obesity) can create edges between diseases that have no direct molecular link. Always validate comorbidity-derived hypotheses against molecular data -- check whether the disease modules overlap or are proximal in the interactome before concluding that two diseases share a mechanism.

Putting It All Together: The Network Medicine Pipeline

The concepts in this chapter form a coherent pipeline for translating molecular data into clinical insight. The pipeline begins with constructing the signaling graph for a pathway of interest, annotating it with disease associations and drug targets, and then applying network analysis to answer clinical questions.

Build the signaling graph: Curate receptor, kinase cascade, second messenger, and transcription factor interactions from databases such as KEGG, Reactome, and SignaLink.
Identify the disease module: Map GWAS-associated genes onto the interactome and extract the connected subgraph using algorithms like DIAMOnD or network propagation.
Prioritize disease genes: Apply guilt-by-association scoring or random walk with restart to rank candidate genes by their proximity to the disease module.
Discover drug targets: Identify proteins within the disease module that are druggable (have binding pockets suitable for small molecules) and validate them using network centrality and genetic evidence.
Repurpose existing drugs: Query the drug-target-disease knowledge graph for drugs that target proteins proximal to the disease module. Use metapath-based scoring (as in Hetionet) to rank candidates.
Predict side effects and interactions: Compute network proximity between the drug's target set and the gene sets of unrelated disease modules to flag potential off-target effects. Check the drug-drug interaction graph for conflicts with the patient's existing medications.
Personalize treatment: Overlay the patient's genomic and transcriptomic data onto the signaling network to identify which modules are active and select therapies targeting those specific modules.

This pipeline illustrates the central promise of network medicine: by integrating diverse data types into a unified graph framework, we can move from identifying disease mechanisms to predicting therapeutic strategies in a systematic, data-driven manner.

Great Work, Investigators!

You have traveled from receptor activation at the cell membrane all the way to precision medicine at the bedside -- and the unifying thread was always the graph. Signaling cascades are directed graphs, diseases are network modules, drug repurposing is a path-finding problem in a knowledge graph, and cancer driver discovery is network propagation on the interactome. These ideas will converge further in the remaining chapters as we layer additional data types onto these network foundations. Onward, explorers!

Key Takeaways

Cell signaling cascades are naturally represented as directed graphs, with nodes for receptors, kinases, second messengers, and transcription factors, and directed edges for activation or inhibition events. Feedback loops and feed-forward loops are recurring motifs that control signal dynamics.
Network medicine treats the human interactome as a map of disease. Disease genes cluster into connected subgraphs called disease modules, and the network proximity between gene sets quantifies their biological relationship.
Guilt-by-association scoring predicts novel disease genes by exploiting the observation that interactome neighbors of known disease genes are themselves likely to be disease-related. This principle underlies disease gene prioritization from GWAS data.
Drug-target-disease knowledge graphs are the computational engine of drug repurposing. Heterogeneous graphs like Hetionet encode relationships among drugs, targets, and diseases, and metapath-based features capture the graph connectivity patterns that predict therapeutic potential.
Cancer driver gene identification requires network context. Mutation frequency alone misses rare drivers; combining mutation data with PPI network propagation reveals genes whose network position makes them functionally important despite low mutation rates.
Precision medicine maps individual patient data onto reference networks to identify activated disease modules and select personalized therapies. Network biomarkers capture pathway-level activity and are more robust than single-gene markers.
Drug safety analysis is a graph problem. Drug-drug interaction graphs, adverse event networks, and side effect prediction via network proximity help manage the risks of polypharmacy and identify off-target effects before they harm patients.
Comorbidity networks reveal hidden relationships between diseases that can be explained mechanistically through overlapping or proximal disease modules in the interactome.

Beyond Molecular Networks: Epidemic Contact Graphs

While this chapter focuses on molecular-level disease networks — signaling cascades, disease modules, and drug-target interactions — graph-based approaches to disease extend well beyond the molecular scale. One compelling example is the epidemic contact graph, where nodes represent individuals in a population and edges represent contacts that may transmit an infectious disease.

Although epidemic contact tracing is not directly related to the core bioinformatics topics of this textbook, it illustrates how the same graph visualization and analysis principles covered throughout this course apply to public health and epidemiology. Contact graphs reveal super-spreader nodes (high-degree hubs), transmission chains (directed paths), and community structure — concepts that parallel hub proteins in PPI networks, signaling cascades, and disease modules.

View Epidemic Contact Graph MicroSim Fullscreen

The simulation above models an SIR (Susceptible-Infected-Recovered) epidemic. Press Start to watch the disease spread through the population. Red arrows show transmission events, while gray dashed lines show contacts that did not result in infection. Adjust the sliders to explore how infection probability, contact rate, and recovery time affect outbreak dynamics and the estimated basic reproduction number (R0).

See Annotated References