Medical knowledge has traditionally been organized like an encyclopedia: alphabetically, by topic, in neat separate volumes. You look up "Breast Cancer" in one place, "BRCA2" in another, "Olaparib" in a third. Each entry is meticulously detailed, scientifically accurate, and completely isolated from the others.
But biology doesn't work like an encyclopedia. Biology is a network.
A disease isn't just a diagnosis—it's a cascade of molecular events involving dozens of genes, proteins, and pathways. A drug doesn't just "treat a condition"—it targets specific proteins, which are encoded by specific genes, which interact with other genes in complex networks. A clinical trial isn't just testing a compound—it's probing a hypothesis about biological relationships.
Traditional medical databases preserve the encyclopedia structure. ClinicalTrials.gov shows you trials. OpenTargets shows you gene-disease associations. PubChem shows you drug structures. Each is excellent at its specific job, but they don't show you the most important thing: how everything connects.
That's why we built Crick's Knowledge Graph—not as another database, but as a map of medical knowledge itself.
At its simplest, a knowledge graph is a way of representing information as entities (nodes) and relationships (edges). Instead of storing data in tables or documents, you store it as a network of interconnected facts.
For Crick, the entities are:
The relationships are what make it powerful:
When you visualize these relationships, something remarkable happens: you see not just data points, but the logical structure of medical knowledge. You understand why certain treatments work, why they work for some patients but not others, and where the gaps and opportunities exist.
Let's compare two approaches to the same question: "What treatments are available for breast cancer patients with BRCA mutations?"
Traditional linear search:
Knowledge Graph exploration:
The difference isn't just efficiency—it's comprehension. The Knowledge Graph shows you why Olaparib makes sense for BRCA2-mutant cancers (both involve DNA repair pathways). You're not blindly following treatment recommendations; you understand the molecular logic.
Let me show you the Knowledge Graph in action with a common clinical scenario.
A patient is diagnosed with non-small cell lung cancer. Their oncologist orders tumor sequencing and finds an EGFR mutation. The patient goes home and searches "lung cancer treatment."
Without a Knowledge Graph, they find information about chemotherapy, radiation, and immunotherapy—generic treatments that apply broadly to lung cancer. They might never discover that their specific genetic mutation makes them ideal candidates for targeted therapy.
With Crick's Knowledge Graph, here's what happens:
Start at "Lung Cancer" node. You see it's connected to numerous genes, including EGFR, KRAS, ALK, ROS1. This immediately tells you that lung cancer isn't one disease—it's many diseases defined by different genetic drivers.
Click on EGFR. The graph shows:
Click on osimertinib. You see:
Now click on one of those trials. You see:
In 10 minutes of visual exploration, the patient understands:
This is knowledge that would take hours of reading research papers to piece together—if you even knew what to search for.
One of the most powerful aspects of knowledge graphs is discovering connections you weren't looking for.
Researchers call this "network topology insight"—understanding that the structure of the network itself contains information.
Hub nodes (entities with many connections) are often druggable targets or common disease drivers. If a gene is connected to dozens of diseases, it's either a fundamental biological process or a promising therapeutic target.
Bridge nodes connect otherwise separate parts of the graph. These are often opportunities for drug repurposing. A drug developed for one disease might work for another if they share a bridge node (a common gene or pathway).
Community clusters are groups of tightly interconnected nodes. These often represent functional modules—genes that work together in a pathway, drugs with similar mechanisms, diseases with shared biology.
Example discovery: A researcher exploring inflammatory bowel disease in the Knowledge Graph notices that several IBD-associated genes are also connected to psoriasis. Clicking through, they discover that some biologics approved for psoriasis are being tested for IBD in early trials. This is drug repurposing discovered through network structure, not literature review.
One of the most exciting aspects of Crick's Knowledge Graph is that it's not static—it evolves as science advances.
When a new clinical trial is registered on ClinicalTrials.gov, it appears in the graph, connected to the disease it's studying, the drug it's testing, and the genetic biomarkers it requires.
When new gene-disease associations are published and incorporated into OpenTargets, new edges appear in the graph.
When FDA approves a new drug, its status changes, and its connections proliferate as more trials test it in combination with other therapies.
This means the Knowledge Graph is a living representation of the current state of medical research. You're not reading a textbook from 2020—you're exploring what researchers are investigating today.
For scientists, this temporal aspect is invaluable. You can see research trends: which targets are "hot" (many new trials), which have cooled off (few recent trials despite many connections), and which are emerging (suddenly many trials after years of dormancy).
Here's where it gets really powerful: personalized knowledge graphs.
When you upload your 23andMe data or enter your specific diagnosis, Crick can highlight the parts of the graph most relevant to you.
Imagine the full Knowledge Graph is a map of the entire world. Personalization is like highlighting your neighborhood, your common routes, places you care about. The full map is still there, but your personal context is emphasized.
A patient with a BRCA2 mutation sees BRCA2-connected nodes emphasized. Trials recruiting BRCA2-positive patients are highlighted. Drugs targeting the DNA repair pathway are flagged. The graph becomes their map, not just a general reference.
This personalization extends to research interests. An oncologist studying immunotherapy can filter the graph to emphasize immune checkpoint inhibitors, CAR-T cells, and related trials. A genetic counselor can focus on hereditary cancer genes and associated conditions.
The same underlying graph serves everyone, but each person navigates it through their own lens.
Crick's Knowledge Graph offers three distinct visualization modes, each revealing different insights:
Ego Mode: Traditional force-directed layout where nodes push apart and edges pull together. Great for seeing immediate connections around a specific entity. When you click a node, it becomes the "ego" at the center with its neighborhood arranged organically around it.
Pathway Mode: Arranges nodes in left-to-right columns by type (Genes → Targets → Diseases → Drugs → Trials). Perfect for understanding biological flow and causality. You can visually trace: gene mutations lead to protein dysfunction, causing disease, which is targeted by drugs being tested in trials.
Cluster Mode: Groups nodes by type in distinct regions with visual "hulls" around each cluster. Excellent for seeing the overall structure and comparing cluster sizes. If you see 50 genes but only 5 drugs in the graph, that reveals an opportunity—many targets, few therapies.
Switch between modes instantly to gain different perspectives on the same data.
The challenge with knowledge graphs is that network visualizations can be overwhelming. Show someone a graph with 10,000 nodes and 50,000 edges, and they'll just see a hairball of lines.
We've invested heavily in making Crick's Knowledge Graph intuitive:
Progressive disclosure: Start with a simple view (disease + immediately connected genes and drugs), then expand as needed by clicking the "+" badges on nodes. You're never confronted with everything at once.
Smart layout: Nodes are positioned using physics simulation that places related entities near each other naturally.
Color coding: Each entity type has a distinct color. Diseases are red circles, genes are green diamonds, drugs are blue hexagons, trials are amber rectangles, targets are teal triangles. Your eye learns the visual language quickly.
Interactive filtering: Control panel lets you toggle entity types on/off. Hide trials, show only Phase 3. Hide genes, show only FDA-approved drugs. The graph adapts instantly.
Click-to-focus: Click any node, and it moves to the center with its immediate neighbors highlighted. The rest of the graph fades to background. This local focus prevents overwhelm while maintaining context.
We've watched non-scientists use the Knowledge Graph with minimal instruction. The key is that it's visual and interactive—you learn by exploring, not by reading a manual.
Perhaps the most important aspect of Crick's Knowledge Graph is what it represents philosophically: the democratization of scientific insight.
For decades, understanding how diseases, genes, and treatments connect required years of medical education and hours of literature review. Knowledge graphs make this insight accessible to anyone willing to click and explore.
A patient newly diagnosed with cancer can gain the molecular understanding that once required a PhD. A medical student can visualize in minutes what would take weeks to learn from textbooks. A researcher in one field can quickly understand another domain by exploring its connections.
This doesn't replace expertise—oncologists, geneticists, and researchers bring irreplaceable knowledge and judgment. But it levels the informational playing field. Patients can have informed conversations with their doctors. Researchers can spot connections across disciplines. The knowledge graph becomes a shared language.
Science is fundamentally about understanding connections: how phenomena relate, how systems interact, how one fact implies another. By representing medical knowledge as a network, we're not just building a tool—we're revealing the actual structure of biomedical science.
And we're making that structure accessible to everyone.
Explore the Knowledge Graph at crick.ai/graph
Discover more articles about precision medicine and biomedical research.
View All Posts