Network Biology and Systems Oncology

Educational note: network models help reason about cancer mechanisms and candidate targets. They are hypothesis engines unless validated experimentally and clinically.

TL;DR

Cancer is not only a list of mutated genes. It is a disturbed system of regulatory circuits, signaling pathways, protein interactions, cell-cell communication, and tissue ecology. Network biology makes those relationships explicit.

The danger is treating graph metrics as clinical truth. A high-degree node, central pathway, or model-predicted control point is not automatically a safe drug target.

Core Network Types

Network	Nodes	Edges	Typical question
Protein interaction	Proteins	Physical or functional interactions	Which proteins sit near a cancer dependency?
Signaling pathway	Receptors, kinases, transcription factors	Directional biochemical relationships	What pathway transmits a growth or survival signal?
Gene regulatory	Transcription factors and genes	Activation, repression, inferred regulation	Which regulators maintain a malignant cell state?
Metabolic	Metabolites, enzymes, reactions	Flux or reaction relationships	Which nutrient or pathway dependencies matter?
Cell-cell communication	Cell types, ligands, receptors	Paracrine or contact interactions	How does the microenvironment shape therapy response?

Common Analyses

• Centrality: identifies connected or influential nodes, but can be biased by literature density.
• Community detection: finds modules; modules need biological annotation.
• Pathway enrichment: summarizes gene lists; sensitive to input quality and background set.
• Network propagation: spreads signal across a graph to prioritize nearby genes or drugs.
• Causal modeling: tries to infer direction, but needs perturbation or strong assumptions.
• Controllability: predicts nodes that may shift a system state; still largely translational in oncology.

Evidence Standards

Before a network claim becomes clinical language, require:

1. A clearly defined biological system and data source.
2. Transparent edge provenance: curated, experimental, inferred, or literature-mined.
3. Sensitivity analysis for graph choice and parameters.
4. Independent validation with perturbation, orthogonal data, or clinical cohort.
5. Separation between target plausibility and druggability.
6. Explicit maturity label: preclinical, translational, clinical, or approved.

Practical Uses

• Prioritizing candidate targets for experiments.
• Explaining resistance pathways and bypass signaling.
• Designing rational combination hypotheses.
• Mapping biomarker panels to pathways.
• Connecting omics results to known biology.

What It Does Not Do

• It does not prove causality from co-expression alone.
• It does not guarantee a target is druggable.
• It does not replace dose, toxicity, pharmacology, or clinical trials.
• It does not make a combination safe just because two nodes appear complementary.

Useful Resources

• Reactome for curated pathways.
• STRING for protein association networks.
• Open Targets Platform for target-disease evidence.
• cBioPortal for cancer genomics context.

References

1. Gillespie M, Jassal B, Stephan R, et al. The Reactome pathway knowledgebase 2022. Nucleic Acids Research. 2022;50(D1):D687-D692. PMID: 34788843. https://doi.org/10.1093/nar/gkab1028
2. Szklarczyk D, Kirsch R, Koutrouli M, et al. The STRING database in 2023. Nucleic Acids Research. 2023;51(D1):D638-D646. PMID: 36370105. https://doi.org/10.1093/nar/gkac1000
3. Ochoa D, Hercules A, Carmona M, et al. Open Targets Platform: supporting systematic drug-target identification and prioritisation. Nucleic Acids Research. 2023;51(D1):D1302-D1310. PMID: 36350656. https://doi.org/10.1093/nar/gkac1046