The Connections Your Doctor Might Not Mention
If you have psoriasis, has anyone told you that your risk of cardiovascular disease is significantly elevated? If you manage type 2 diabetes, did your endocrinologist mention the growing research linking it to Alzheimer's disease? If you live with inflammatory bowel disease, were you warned about the connection to joint inflammation?
These are not fringe theories. They are well-documented disease relationships supported by decades of research. But the sheer volume of medical knowledge — over 30 million papers indexed in PubMed alone — makes it nearly impossible for any single clinician to track every meaningful connection between conditions.
This is where disease knowledge graph connections become genuinely useful. Not as a replacement for clinical expertise, but as a structured way to surface relationships that matter to patients navigating complex health situations.
What Is a Disease Knowledge Graph?
A knowledge graph is a structured database that maps entities (things) and the relationships between them. Unlike a flat list or a simple table, a knowledge graph captures how concepts connect to each other in multiple dimensions.
A disease knowledge graph specifically maps how diseases relate to genes, proteins, drugs, biological pathways, anatomical structures, and — critically — other diseases. Think of it as a map of medical knowledge where every node is a biological entity and every line between nodes represents a documented relationship.
This matters because diseases do not exist in isolation within the human body. They share genetic risk factors, affect overlapping biological pathways, respond to similar drugs, and co-occur in patients far more often than chance would predict. A knowledge graph makes these connections explicit and queryable.
For a deeper introduction to how knowledge graphs work in healthcare contexts, see our overview of knowledge graphs in health AI.
How PrimeKG Maps Disease Connections
PrimeKG (Precision Medicine Knowledge Graph) is one of the most comprehensive biomedical knowledge graphs available for research. Published by Chandak et al. in Nature Scientific Data in 2023, it integrates data from 20 high-quality biomedical resources into a single, unified structure (Chandak, P., Huang, K., & Zitnik, M. "Building a knowledge graph to enable precision medicine." Scientific Data, 10, 67, 2023).
The numbers give a sense of its scope:
- 17,080 diseases mapped with their relationships
- 29,786 genes linked to disease mechanisms
- 4,050 drugs connected to the conditions they treat and their molecular targets
- Over 4 million relationships connecting these entities across 10 types of nodes
What makes PrimeKG particularly relevant for patient support is its inclusion of disease-disease relationships, drug-disease connections, and the biological pathways that explain why certain conditions tend to cluster together. For more detail on PrimeKG's construction and what it contains, see What Is PrimeKG?.
Surprising Disease Knowledge Graph Connections
Here are several examples of disease relationships that a knowledge graph can surface — connections that are well-supported by research but may not be obvious to patients managing one of these conditions.
Psoriasis and Cardiovascular Disease
Psoriasis is often perceived as "just a skin condition." In reality, the chronic systemic inflammation that drives psoriasis also accelerates atherosclerosis. Multiple large cohort studies have found that patients with severe psoriasis have a significantly elevated risk of myocardial infarction, stroke, and cardiovascular mortality.
In a knowledge graph, this connection is visible through shared inflammatory pathways — particularly involving TNF-alpha, IL-6, and IL-17 signaling — that drive both psoriatic plaques and arterial plaque formation. The connection is strong enough that some cardiologists now consider severe psoriasis an independent cardiovascular risk factor.
Type 2 Diabetes and Alzheimer's Disease
The link between type 2 diabetes and Alzheimer's disease has become increasingly well-established. Epidemiological studies show that people with type 2 diabetes have roughly a 60% greater risk of developing dementia. The shared mechanisms include insulin resistance in brain tissue, chronic inflammation, and vascular damage.
A knowledge graph makes this visible by showing overlapping genetic risk factors (genes like APOE and insulin signaling genes), shared metabolic pathways, and the biological mechanisms connecting glucose metabolism to amyloid protein accumulation. Some researchers have gone so far as to describe Alzheimer's as "type 3 diabetes," though this framing remains debated.
Depression and Heart Disease
The bidirectional relationship between depression and cardiovascular disease is one of the most studied comorbidity connections in medicine. Depression increases cardiovascular risk through multiple mechanisms: elevated cortisol, increased inflammatory markers, platelet activation, and behavioral factors like reduced physical activity and medication non-adherence.
In PrimeKG, this connection appears through shared genes (particularly those involved in serotonin signaling and the HPA axis), overlapping drug targets (SSRIs have demonstrated antiplatelet effects), and documented disease-disease co-occurrence patterns.
Inflammatory Bowel Disease and Arthritis
Up to 30% of patients with inflammatory bowel disease (Crohn's disease or ulcerative colitis) develop some form of arthritis. This is not coincidence — both conditions involve dysregulation of the same immune pathways, particularly those involving IL-23, Th17 cells, and gut-joint axis signaling.
Knowledge graphs capture this through shared genetic susceptibility loci (including IL23R variants), common drug targets (TNF inhibitors treat both conditions), and documented anatomical pathway connections between gut mucosa and synovial tissue.
For more on how co-occurring conditions affect patient experience, see our guide on understanding comorbidities and multiple conditions.
Why This Matters for Patient Support
Disease knowledge graph connections have direct implications for how patients find support and information.
Connecting Patients Across Conditions
When a knowledge graph reveals that psoriasis and cardiovascular disease share biological mechanisms, it suggests that patients managing both conditions may benefit from connecting with each other — not just within siloed condition-specific support groups, but across them.
A patient newly diagnosed with cardiovascular disease who also has psoriasis may gain useful perspective from others who have navigated that specific combination. Traditional patient support groups are typically organized around single conditions, but comorbidity-aware tools can help bridge those gaps.
Understanding Your Full Picture
Many patients manage multiple conditions simultaneously. Knowledge graphs help explain why certain conditions cluster together, which can reduce the confusion and frustration of accumulating diagnoses. Understanding that your conditions share underlying mechanisms — rather than being random bad luck — can be genuinely reassuring and can inform better conversations with your care team.
For evidence on how peer support affects health outcomes, see the research on peer support for chronic illness.
Identifying Relevant Information
When you search for health information online, you are competing with an enormous amount of low-quality content. Knowledge graphs provide a structured, source-verified alternative to unfiltered search results. Rather than relying on SEO-optimized articles of uncertain quality, a knowledge graph draws from peer-reviewed databases with documented provenance.
This does not eliminate the problem of health misinformation online, but it provides a more reliable foundation for AI systems that help patients find relevant information.
How AI Uses Knowledge Graphs vs. Generic Training Data
Large language models (LLMs) like those powering many AI health tools are trained on vast amounts of internet text. This gives them broad knowledge but also introduces significant problems: they can "hallucinate" plausible-sounding but incorrect medical information, they cannot cite specific sources for their claims, and their knowledge has a training cutoff date.
Knowledge graphs address these problems through a technique called retrieval-augmented generation (RAG). Instead of relying solely on what the model "remembers" from training, the system queries a structured knowledge graph for relevant facts and relationships, then uses those verified facts to generate its response.
A 2024 systematic review by Li et al. found that knowledge graph-augmented approaches significantly improved the factual accuracy of AI systems in biomedical question-answering tasks, particularly for questions involving multi-hop reasoning across disease relationships (Li, X., et al. "Knowledge Graphs for Biomedical Natural Language Processing: A Review." Journal of Biomedical Informatics, 149, 104577, 2024).
PatientSupport.AI uses this approach — combining PrimeKG's structured disease knowledge graph with Groq Llama 70B to provide AI-assisted health information that is grounded in documented relationships rather than pattern-matched from training text. The tool is free to use without creating an account.
The practical difference: when you ask about connections between two conditions, the system can trace a specific path through the knowledge graph and explain the biological basis for the connection, rather than generating a generic response that may or may not be accurate.
Limitations and Honest Caveats
Intellectual honesty requires acknowledging what knowledge graphs cannot do.
They Are Not Complete
PrimeKG contains 17,080 diseases, but the full spectrum of human disease is larger and constantly being refined. Rare diseases, newly characterized conditions, and conditions with limited research may be poorly represented or absent entirely. The graph is only as good as the databases it draws from.
Knowledge Graphs Reduce But Do Not Eliminate AI Errors
Grounding AI responses in a knowledge graph significantly reduces hallucination compared to pure language model generation. But it does not eliminate errors entirely. The AI may still misinterpret queries, retrieve partially relevant information, or generate responses that oversimplify complex relationships. Any health information from an AI system — including ours — should be discussed with a qualified healthcare provider before acting on it.
Correlation and Mechanism Are Not Treatment Guidance
A knowledge graph can show that two diseases share genetic risk factors or biological pathways. It cannot tell you whether a specific treatment for one condition will help with another. Drug repurposing based on knowledge graph connections is an active area of research, but it requires rigorous clinical trials before becoming clinical practice. Do not adjust your treatment based on disease connections alone.
Biological Relationships Are Not Deterministic
Having one condition that is connected to another in a knowledge graph does not mean you will develop the second condition. These are population-level statistical associations and shared biological mechanisms, not individual predictions. Many people with psoriasis never develop heart disease. Many people with diabetes never develop Alzheimer's.
Using Disease Knowledge Graph Connections Responsibly
Disease knowledge graphs represent a meaningful advance in how we organize and access medical knowledge. They make explicit the connections between conditions that have traditionally been scattered across thousands of individual research papers. For patients managing complex health situations, this can provide context, reduce confusion, and support more informed conversations with healthcare providers.
But they are tools for understanding, not tools for self-diagnosis or self-treatment. The value of disease knowledge graph connections lies in helping you ask better questions — of your doctors, of your support networks, and of the AI systems that increasingly mediate access to health information.
If you want to explore how conditions relate to each other through documented biological connections, PatientSupport.AI makes PrimeKG's knowledge graph accessible through natural conversation — free, without requiring an account, and grounded in peer-reviewed data sources rather than unverified internet content.
Patient support groups and AI health tools are complements to professional medical care, not replacements. If you are experiencing a medical emergency, contact your healthcare provider or call 911.