Name the question
Bad: “What causes lupus?” Better: “What mechanisms are studied in lupus-related kidney injury?”
Understand the medicine. Then investigate the science.
EvidenceHelix explains biomedical concepts twice: first in accurate scientific language, then again in plain language. Use the Medical Book to learn the biology, and the Scientist Handbook to evaluate evidence, design experiments, challenge assumptions, and move findings into the Discovery Bench.
Medical Bookdisease, genes, drugs, environment, population
Scientist Handbookevidence, experiments, cohorts, reproducibility
Visual firstcharts, pathways, comparison tables, diagrams
Governedsource rights, provenance, uncertainty, safety
Teaching charts are labeled as conceptual unless they are tied to a cited dataset. EvidenceHelix is for research and education, not diagnosis or treatment selection.
No chapter matched that search yet. Try a broader word such as gene, cancer, environment, trial, or population.
Your first 30 minutes in EvidenceHelix
You do not need a medical degree to start asking good scientific questions. The trick is to turn a broad concern into a narrow, checkable research question.
Define the terms
Look up the disease, organ, biomarker, gene, medication, or exposure before comparing studies.
Collect 3 evidence types
Start with an authoritative overview, then a review/article, then a trial, dataset, genetics record, or regulatory source.
Build an evidence table
For each claim record the population, design, outcome, date, source, and what the evidence does not prove.
Look for disagreement
Ask whether differences come from population, endpoint, dose/exposure, study design, follow-up, or confounding.
Move to Discovery Bench
Only after the source trail is visible should you draft a candidate hypothesis and a possible falsifying experiment.
Beginner research loop
Question→
Definitions→
Evidence table→
Contradictions→
Hypothesis→
Test / reproduce
Try this first:
How can lupus affect kidney function, and what evidence distinguishes lupus nephritis from other causes of CKD?
That single question naturally teaches disease biology, kidney function, labs, pathology, study design, differential evidence, and source provenance—without asking the software to diagnose a person.
Disease chapters: from name → mechanism → research questions
These starter chapters are deliberately structured the same way so the reader can compare diseases without pretending they all work alike.
Sickle cell disease
Scientific: an inherited hemoglobin disorder caused by pathogenic variants in HBB that produce hemoglobin S and can alter red-cell shape, rheology, oxygen delivery, and vascular behavior.
Plain language: a change in the gene for part of hemoglobin can make some red blood cells become rigid or sickle-shaped, which can block blood flow and damage tissues.
HBBhemoglobinred blood cellvascular
- Ask: which findings are genotype-specific?
- Ask: what outcomes were measured—pain, hospitalization, organ damage, hemolysis?
- Separate approved therapies from experimental strategies.
Type 2 diabetes
Scientific: a heterogeneous metabolic disorder involving insulin resistance, impaired beta-cell function, altered glucose regulation, and multi-organ effects.
Plain language: the body has increasing difficulty using insulin effectively and keeping blood sugar in a healthy range.
insulinbeta cellglucosemetabolism
- Population context matters strongly.
- Diet, sleep, medications, genetics, and environment can be relevant research dimensions.
- Short-term glucose measures and long-term outcomes are not interchangeable.
Breast cancer
Scientific: a family of malignant breast tumors with clinically important variation in hormone-receptor status, HER2 status, genomic features, grade, stage, and treatment response.
Plain language: “breast cancer” is not one single disease. Different tumors can depend on different biological signals and behave differently.
ER/PRHER2stagegrade
- Always identify subtype before comparing studies.
- Stage describes spread; grade describes how abnormal/aggressive the cells appear.
- Biomarker status can change the research question.
Glioblastoma
Scientific: an aggressive diffuse glioma defined using modern histologic and molecular criteria; genomic and epigenomic context can materially change interpretation.
Plain language: a fast-growing brain tumor whose exact molecular features matter when scientists compare cases or experiments.
brainIDHgenomicstumor microenvironment
- Do not compare old and new diagnostic labels without checking classification criteria.
- Cell-line, organoid, animal, and human evidence answer different questions.
Systemic lupus erythematosus
Scientific: a systemic autoimmune disease with heterogeneous immune dysregulation and variable involvement of skin, joints, kidneys, nervous system, blood cells, and other organs.
Plain language: the immune system can mistakenly attack the body in different ways, so two people with lupus may have very different patterns.
autoimmunityantibodieskidneyinflammation
- Phenotype definition is critical.
- Compare flare activity, organ involvement, treatment context, and ancestry/population descriptors.
Alzheimer disease
Scientific: a progressive neurodegenerative disease associated with characteristic neuropathology including amyloid and tau changes, while genetics, vascular biology, immunity, aging, and other mechanisms remain active research areas.
Plain language: memory and thinking decline as brain biology changes over time; scientists study several interacting mechanisms rather than one simple cause.
amyloidtauagingcognition
- Biomarker-positive disease and clinical symptoms are related but not identical concepts.
- Longitudinal evidence is especially important.
Cystic fibrosis
Scientific: an autosomal-recessive disorder caused by pathogenic variants in CFTR, affecting chloride and bicarbonate transport and multiple organs.
Plain language: changes in the CFTR gene disrupt a channel that helps move salt and water, which can make mucus unusually thick in organs such as the lungs.
CFTRion channellungpancreas
- Variant class can matter for mechanism-focused research.
- Gene/protein structure links are useful for visualizing why different variants behave differently.
Major depressive disorder
Scientific: a heterogeneous psychiatric syndrome defined clinically by patterns of mood, cognitive, physical, and functional symptoms; biology is multifactorial and no single biomarker explains all cases.
Plain language: depression can affect mood, sleep, thinking, energy, appetite, and function, and researchers study many biological and environmental pathways.
moodsleepcognitionenvironment
- Symptom scales, diagnoses, and biological measurements are different kinds of evidence.
- Avoid treating genetic association as deterministic prediction.
Schizophrenia
Scientific: a heterogeneous psychiatric disorder involving psychotic, cognitive, negative, and functional symptoms with complex genetic and environmental contributions.
Plain language: schizophrenia can affect perception, thinking, motivation, and daily function; its biology is complex and cannot be reduced to one gene or one chemical.
psychosiscognitionpolygenicneurodevelopment
- Medication response and side-effect evidence need separate analysis.
- Population, age, substance exposure, sleep, and physical-health context can matter.
Human Atlas 3D: spin the body, isolate systems, then follow the evidence
The Human Atlas is a dedicated full-page EvidenceHelix lab at
/app/labs/visual-3d-lab. The production viewer renders real HuBMAP Human
Reference Atlas (HRA) GLB anatomy under CC BY 4.0. It keeps the realistic medical preview
visible while the high-detail mesh loads and fails closed rather than substituting fake
organ circles when a 3D asset is unavailable.
Trans-inclusive anatomy
Reference views are not identity-to-anatomy rules.
Male, female, trans male, and trans female views are research presets. Gender identity does
not determine a single set of organs, hormones, surgeries, chromosomes, or body traits.
Anatomy and care overlays must remain independently controllable.
Move around the body
- Drag for 360° rotation.
- Use the wheel or controls to zoom in and out.
- Enable auto-spin or reset the camera.
- Click a visible structure to identify and focus it.
- Turn labels on or off.
Isolate body systems
- Major organs.
- Blood / circulation and major-vessel teaching paths.
- Muscles and skeleton.
- Endocrine structures including the thyroid and pituitary.
- Reproductive anatomy with individual-variation caveats.
Trans-inclusive care overlays
Trans reference views can expose educational research regions for chest masculinization, hysterectomy, phalloplasty, metoidioplasty, breast augmentation, vaginoplasty, orchiectomy, facial gender-affirming surgery, and endocrine context.
These overlays locate an evidence topic. They do not show operative steps, prescribe a technique, promise an outcome, or imply that a trans person has had or wants surgery.
Structure → evidence
A selected structure should become a research anchor for literature, provenance, contradictions, genes, biomarkers, pathways, public clinical trials, multi-omics, cohorts, hypotheses, and reproducibility.
Example: thyroid → TSH/T3/T4 → genes/pathways → literature → trials → cohort context → contradictions.
Current anatomy foundation
| Layer | Current examples | Research direction |
|---|---|---|
| Organs | Brain, heart, lungs, liver, stomach, kidneys, intestines, bladder. | Disease overlays, biomarkers, tissue evidence. |
| Endocrine | Pituitary, thyroid, adrenal glands, pancreas, gonadal context. | Hormone pathways, biomarkers, pharmacology, cohorts. |
| Circulation | Major arterial/venous teaching paths. | Coronary, cerebral, renal, pulmonary and systemic circulation. |
| Musculoskeletal | Pectoral muscles, quadriceps, skull, spine, pelvis and major skeletal paths. | Named high-fidelity bones, muscles, joints, tendons and fascia. |
| Reproductive | Uterus/ovaries and prostate/testes where applicable. | Independent structure toggles; do not infer anatomy from identity. |
| Care / surgery overlays | Research regions for gender-affirming care. | Outcome, complication, follow-up and population-context evidence. |
3D model gate: HRA is approved under CC BY 4.0 with visible attribution. Every
additional external 3D asset still needs documented owner, source, commercial-use permission,
attribution, redistribution/modification rights, provenance, and
ShareAlike/copyright review. Unknown or incompatible rights fail closed.
Teaching atlas — not a surgical planner.
The model is for biomedical research and education, not diagnosis, personalized anatomy,
treatment selection, operative instruction, or outcome prediction.
Conditions can overlap without having one simple cause
This map is a teaching model for organizing research. An arrow means “scientists study a relationship here,” not “this definitely caused that in a particular person.”
Immune systemautoantibodies · inflammation · cytokines
Kidneysfiltration · eGFR · albuminuria · glomeruli
Pain + sensory processingpain sensitivity · fatigue · sleep
Mind + brainmood · cognition · sleep · stress biology
Environment + contextexposure · behavior · access · social factors
Lupus ↔ immune inflammation ↔ kidney involvement
Diabetes ↔ glucose/vascular biology ↔ kidney risk
Fibromyalgia ↔ pain sensitivity ↔ sleep/fatigue/cognition
CKD ↔ filtration/albumin loss ↔ cardiovascular/metabolic context
Mental health ↔ sleep/cognition/stress ↔ physical comorbidity
Cross-condition research matrix
| Condition | Core biology to learn first | Common research measurements | Important “don’t assume” |
|---|---|---|---|
| Lupus (SLE) | Autoimmunity, inflammation, organ involvement, flare/remission. | Symptoms, autoantibodies, CBC, kidney labs/urine, organ-specific measures. | A positive ANA alone does not diagnose lupus. |
| Fibromyalgia | Chronic widespread pain/tenderness, increased pain sensitivity, sleep and fatigue biology. | Pain distribution, fatigue, sleep, cognition, function. | Pain severity does not map neatly to a single tissue-damage marker. |
| Chronic kidney disease | Persistent kidney damage or reduced filtration. | eGFR, creatinine, UACR/albuminuria, urine findings, blood pressure. | CKD in a person with diabetes is not automatically diabetic kidney disease. |
| Diabetes | Glucose regulation and insulin biology; type matters. | Glucose measures, A1C, complications, kidney/eye/nerve/cardiovascular outcomes. | Type 1 and type 2 have different underlying mechanisms. |
CKD, lupus nephritis, and diabetes: learn the measurements before interpreting the story
Chronic kidney disease (CKD)
Scientific: persistent abnormalities of kidney structure or function can reduce filtration and/or allow substances such as albumin to appear abnormally in urine.
Plain language: the kidneys are damaged or are not filtering blood the way they should. Early CKD can have no obvious symptoms.
Beginner measurements: eGFR estimates filtration; UACR looks for albumin leaking into urine.
Lupus nephritis
Scientific: kidney disease caused by systemic lupus erythematosus, where autoimmune inflammation can affect glomeruli and other kidney structures.
Plain language: lupus can sometimes involve the kidneys. Blood tests, urine tests, and sometimes kidney biopsy help researchers and clinicians understand the pattern.
Fibromyalgia
Scientific: a chronic disorder involving widespread pain and tenderness, fatigue, sleep problems, and increased sensitivity to pain; the cause is not fully understood.
Plain language: the nervous system appears to process pain differently, and sleep, fatigue, memory/concentration, and other symptoms can be part of the picture.
Diabetes
Scientific: diabetes involves chronically elevated blood glucose because insulin production and/or insulin action is disrupted; the mechanism differs by diabetes type.
Plain language: glucose stays in the blood instead of being regulated normally. Over time, high glucose can damage organs including the kidneys, nerves, eyes, and heart.
CKD filtration categories
G1≥90normal/high*
G260–89mildly decreased*
G3a45–59mild–moderate
G3b30–44moderate–severe
G415–29severely decreased
G5<15kidney failure range
eGFR in mL/min/1.73 m². *G1/G2 only meet CKD criteria when other markers of kidney damage are present. Classification is educational and must be interpreted in clinical context.
Diabetes types at a glance
| Type 1 diabetes | Type 2 diabetes | Gestational diabetes | |
|---|---|---|---|
| Core mechanism | Immune-mediated destruction of insulin-producing beta cells; little/no insulin production. | Insulin resistance plus inadequate insulin production over time. | Diabetes first diagnosed during pregnancy. |
| Research lens | Autoimmunity, beta cells, genetics, insulin replacement, technology. | Metabolism, beta-cell function, genetics, environment, complications. | Pregnancy physiology, maternal/child outcomes, later metabolic risk. |
| Kidney relevance | Long-term high blood glucose can contribute to kidney damage; kidney disease still requires its own evaluation. | ||
Important: NIDDK specifically cautions researchers not to assume that CKD in someone with diabetes is automatically diabetic kidney disease; other causes can coexist.
DNA → variant → RNA → protein → phenotype
A useful mental model
DNA→
Gene→
RNA→
Protein→
Cell / tissue effect→
Phenotype
Real biology branches, loops, regulates, and compensates; this is a teaching diagram, not a complete molecular pathway.
Variant classification is not the same as “how sick someone will be”
Pathogenic
Likely pathogenic
VUS
Likely benign
Benign
| Term | What it means | Common mistake to avoid |
|---|---|---|
| Pathogenic | Evidence supports a disease-causing role in a defined gene-disease context. | It does not by itself predict severity, age of onset, or treatment response. |
| Likely pathogenic | Evidence strongly leans disease-causing but is not at the highest certainty category. | Do not silently convert “likely” into certainty. |
| VUS | Variant of uncertain significance; current evidence is insufficient or conflicting. | Do not treat a VUS as a confirmed diagnosis. |
| Likely benign / benign | Evidence argues against the variant being disease-causing in that context. | Other variants, genes, or mechanisms can still matter. |
Germline vs somatic
| Germline | Somatic | |
|---|---|---|
| Where it comes from | Present in egg/sperm lineage and typically across many body cells. | Acquired in a subset of cells during life. |
| Often relevant to | Inherited disease risk, family studies, reproductive genetics. | Cancer, clonal cell populations, tissue-specific processes. |
| Family interpretation | May have implications for relatives. | Usually not inherited in the same way. |
EvidenceHelix should always show the gene, transcript/reference sequence, variant notation, classification source, review status, date, and disease context together.
Biomarkers answer different questions
Diagnostic
Helps identify or confirm a disease or biological state.
“Does this pattern support the diagnosis?”
Prognostic
Associated with future disease course or outcome independent of a specific treatment.
“What might happen over time?”
Predictive
Associated with differential response to a particular intervention.
“Who may respond differently to this treatment?”
Pharmacodynamic / monitoring
Changes in response to treatment or tracks biological status over time.
“Is biology changing after an intervention?”
A statistically associated marker is not automatically clinically useful. Analytical validity, clinical validity, and clinical utility are separate questions.
The drug journey: target → evidence → use
Biological target→
Preclinical evidence→
Clinical trials→
Regulatory review→
Label + safety monitoring
| Evidence object | What it can tell you | What it cannot prove alone |
|---|---|---|
| Mechanism / target | How a drug is expected to interact with biology. | That it will improve patient outcomes. |
| Clinical trial | Effects under a defined protocol and population. | That every patient will respond similarly. |
| FDA/official label | Approved indications, dosing framework, contraindications, warnings. | All future evidence or every off-label research question. |
| FAERS / spontaneous reports | Potential post-market safety signals. | Causation or incidence rates by itself. |
| Pharmacogenomics | Gene-drug relationships that may affect metabolism, efficacy, or safety. | That genetics is the only determinant of response. |
Health is not only genes and medications
Exposome is a broad research idea covering environmental exposures across life—chemical, physical, social, behavioral, and biological.
Sourcetraffic, workplace, product, water, food, wildfire, housing
Exposureinhaled, ingested, skin contact, measured or modeled
Internal biologymetabolism, inflammation, endocrine signaling, oxidative stress
Outcome researchsymptoms, biomarkers, disease association, longitudinal outcomes
| Evidence type | Strength / use | Important limitation |
|---|---|---|
| Personal measurement | Direct measure for a person/sample/time window. | May miss earlier or intermittent exposure. |
| Environmental monitor | Measures air/water/environment at a location. | Not identical to individual dose. |
| Modeled exposure | Estimates exposure where direct measurement is unavailable. | Depends on model assumptions and spatial/temporal resolution. |
| Ecological association | Useful for population-level patterns. | Cannot be assumed to describe individual causation. |
| Toxicology / bioassay | Explores biological activity and plausible mechanisms. | Experimental dose/context may differ from real-world exposure. |
The Environment Lab should always label measured vs modeled vs reported vs ecological exposure evidence.
Who, where, how often—and compared with what?
| Term | Plain-language meaning | Typical denominator |
|---|---|---|
| Incidence | New cases appearing during a period. | People at risk over time. |
| Prevalence | How many people have the condition at a point/period. | Total defined population. |
| Mortality rate | Deaths in a defined population/time. | Population and time interval. |
| Relative risk | Risk in one group divided by risk in another. | Comparison groups. |
| Absolute risk difference | The actual difference in probability between groups. | Same defined time horizon. |
Sensitivity vs specificity
| Sensitivity | Specificity | |
|---|---|---|
| Question | Among people who truly have the condition, how many test positive? | Among people who truly do not have it, how many test negative? |
| High value helps | Reduce false negatives. | Reduce false positives. |
| Does not determine alone | Positive predictive value, negative predictive value, and real-world usefulness also depend on prevalence and context. | |
Population relevance checklist
The bars are a checklist visualization, not a scientific score.
Health equity lens: always ask who was underrepresented, excluded, lost to follow-up, or missing from the dataset—and whether a result is being generalized beyond the population actually studied.
Mental health is not an appendix to medicine
EvidenceHelix treats psychiatric research as first-class biomedical science: clinical phenotype, cognition, sleep, genetics, brain/circuit research, medication evidence, environment, physical comorbidity, function, and lived context. No condition is reduced to one gene or one neurotransmitter.
Mooddepression · bipolar
Anxiety / threatGAD · panic · PTSD
Psychosisschizophrenia spectrum · mood disorders with psychosis
CompulsivityOCD · repetitive behavior
Cognitionattention · memory · processing · executive function
Sleep / circadiansleep duration · timing · disruption · bidirectional effects
Major depressive disorder
Depression can involve persistent low mood or loss of interest plus changes in sleep, energy, cognition, appetite, movement, and function. NIMH describes genetic, biological, environmental, and psychological contributors.
Beginner question: which symptom domain or biological mechanism is a study actually measuring?
Anxiety disorders
Anxiety disorders involve more than ordinary worry or fear and can interfere with daily life. Different disorders have different triggers and symptom patterns.
Beginner question: is the paper studying diagnosis, symptom severity, avoidance, physiology, or treatment response?
Bipolar disorder
Bipolar disorders involve episodes with marked changes in mood, energy, activity, and concentration, including mania/hypomania and depressive episodes.
Beginner question: which episode type and time period does the evidence refer to?
Schizophrenia
Research commonly separates psychotic, negative, cognitive, functional, developmental, genetic, and environmental dimensions rather than treating schizophrenia as a single measurable feature.
Beginner question: which symptom domain, stage of illness, treatment context, and population were studied?
Obsessive-compulsive disorder (OCD)
OCD involves recurring intrusive thoughts/urges/images (obsessions), repetitive behaviors or mental acts (compulsions), or both, with significant distress or interference.
Beginner question: does the study measure obsessions, compulsions, impairment, treatment response, or a proposed mechanism?
Post-traumatic stress disorder (PTSD)
PTSD can develop after traumatic exposure and is studied across intrusion/memory, avoidance, mood/cognition, arousal, sleep, stress biology, and functioning.
Beginner question: what trauma exposure, time since exposure, symptom domain, and comparison group are defined?
Mental-health evidence matrix
| Research layer | Examples | What to be careful about |
|---|---|---|
| Clinical phenotype | diagnostic interviews, symptom scales, functioning | A scale score is not the whole person or automatically a diagnosis. |
| Cognition | memory, attention, processing speed, executive tasks | Task performance depends on context, sleep, medication, practice, and many other factors. |
| Genetics | GWAS, polygenic signals, rare variants | Association is not destiny; no single common gene explains a complex psychiatric disorder. |
| Neurobiology | electrophysiology, circuits, molecular studies | Group-average brain differences are not individual diagnostic tests. |
| Medication | efficacy, adverse effects, pharmacogenomics, adherence | Average trial effects do not predict every individual's response. |
| Environment/context | trauma, stress, sleep, substance exposure, neighborhood/social factors | Environmental association does not automatically establish causation. |
Historical U.S. survey estimates — context matters
Past-year U.S. adult estimates reported by NIMH from the National Comorbidity Survey Replication; these underlying survey data are historical and are shown here to teach how prevalence charts should be labeled, not as a claim about 2026 prevalence.
Words you will see all over the lab
PhenotypeObservable traits, symptoms, measurements, or disease features.
GenotypeThe genetic makeup or specific genetic variants being considered.
PenetranceHow often a genotype is associated with the expected phenotype.
ExpressivityHow much the phenotype varies among people with a relevant genotype.
StageExtent/spread of disease, especially in cancer.
GradeMicroscopic or biological appearance/aggressiveness; not the same as stage.
EndpointThe specific outcome a study was designed to measure.
Surrogate endpointA substitute measure expected to predict a clinically meaningful outcome.
ConfounderA third factor associated with both exposure and outcome that can distort an observed relationship.
Hazard ratioA time-to-event comparison between groups; not the same as absolute risk.
Odds ratioA ratio of odds; can differ substantially from risk ratio when outcomes are common.
Confidence intervalA range reflecting uncertainty around an estimate under a statistical model.
Do not flatten every paper into “proof”
Conceptual evidence ladder
Synthesis / well-conducted systematic review
Randomized or strong interventional evidence
Prospective / observational human evidence
Case reports / case series
Preclinical / in vitro / computational hypotheses
This is a teaching framework, not a universal ranking. Quality within each design can vary dramatically.
Correlation vs causation
| Observation | What it supports | What is still needed |
|---|---|---|
| Exposure X is more common in people with outcome Y. | An association worth investigating. | Temporal order, confounder control, replication, dose-response, mechanistic plausibility, stronger designs where possible. |
| Gene variant correlates with phenotype. | Genetic association. | Replication, functional evidence, population context, segregation where relevant, careful causal inference. |
| Adverse event appears after a medication. | A temporal signal. | Background rate, dechallenge/rechallenge context, alternative causes, denominator, controlled evidence. |
EvidenceHelix should say associated with when that is what the data show—and reserve causal language for evidence that actually supports it.
Turn an interesting pattern into a testable study
Observation→
Hypothesis→
Prediction→
Experiment→
Analysis→
Reproduce / challenge
| Design item | Question to write down before running |
|---|---|
| Hypothesis | What specific, falsifiable relationship are we testing? |
| Population / model | Which people, cells, animals, tissues, datasets, or samples does this claim apply to? |
| Comparison | What is the control/reference group? |
| Outcome | What exact measurement determines whether the prediction held? |
| Confounders | What else could create the same pattern? |
| Missing data | How will missingness be handled? |
| Analysis plan | Which analysis is primary, and which are exploratory? |
| Replication | How could another team or dataset reproduce the result? |
Different omics layers describe different parts of biology
| Layer | What it measures | Typical research question |
|---|---|---|
| Genomics | DNA sequence and variation. | Which variants are associated with disease or function? |
| Epigenomics | Regulatory marks and chromatin state. | Which regions are active, repressed, or differently regulated? |
| Transcriptomics | RNA expression. | Which genes are being transcribed, in which cells or conditions? |
| Proteomics | Protein abundance/modification. | Which proteins or pathways are changing? |
| Metabolomics | Small molecules and metabolic products. | Which biochemical processes appear altered? |
| Single-cell / spatial | Cell-level identity and location. | Which cell populations drive the signal, and where? |
Cross-omics agreement can strengthen a hypothesis, but correlated layers are not independent proof.
A cohort is defined by who got in—and who did not
Eligibility
Age, diagnosis criteria, disease subtype, prior treatment, labs, geography, ancestry descriptors, comorbidities.
Follow-up
How long participants were observed and how many were lost to follow-up.
Missingness
Which measurements are absent, whether absence is systematic, and how analyses handle it.
External validity
Whether results plausibly generalize beyond the people or samples actually studied.
EvidenceHelix should display a population-match explanation, not a fake “relevance percentage.”
Two studies can disagree without one being “wrong”
| Difference to inspect | Why it can change the result |
|---|---|
| Population | Age, ancestry, disease severity, subtype, comorbidities, prior treatment. |
| Design | Randomized vs observational vs case series vs preclinical model. |
| Dose / exposure | Different amounts, duration, route, timing, or real-world exposure. |
| Endpoint | Biomarker change may not equal symptom or survival benefit. |
| Follow-up | Short studies may miss long-term benefit or harm. |
| Bias / confounding | Selection, measurement, publication, and residual confounding can shift estimates. |
Where findings become candidate discoveries
The Discovery Bench should never produce “magic certainty.” It assembles governed evidence into a transparent candidate explanation.
Findings+
Irregularities+
Contradictions+
Omics / environment / cohorts→
Candidate hypothesis
Scientific view
Claim, supporting evidence, contradicting evidence, confounders, source provenance, population, mechanisms, falsifiers, next experiments.
Plain-language view
“Here is the pattern we found, why it might matter, what does not fit yet, and what would need to be tested next.”
Can another scientist retrace the work?
Data provenance
Dataset/version, query, filters, source IDs, retrieval date, licensing decision.
Method provenance
Code version, environment, package versions, random seeds where relevant, analysis parameters.
Notebook capsule
Jupyter notebook, inputs, outputs, figures, assumptions, unresolved warnings.
Challenge log
Contradictions reviewed, alternative explanations, failed analyses, negative results, changes to the hypothesis.
Publicly visible does not automatically mean reusable
| Decision | Meaning | System behavior |
|---|---|---|
| ALLOW | Terms support the declared production use. | Connector can ingest while preserving required attribution/provenance. |
| CONDITIONAL | Use is possible only after a documented condition is satisfied. | Fail closed until condition is proven. |
| DENY | Terms are incompatible or rights are unknown. | No production ingestion. |
| EXTERNAL ONLY | Useful research tool, but not a safe core-data dependency. | Link out; do not scrape or mirror. |
Default = DENY. Unknown licensing never becomes “probably okay.”
How the product pieces fit together
Disease Atlas + Human Atlas 3D
Disease concepts plus a rotatable, zoomable anatomy lab with named body-system layers and structure-to-evidence research paths.
Mind & Brain
Psychiatry, neurology, cognition, sleep, medication, genetics, environment, comorbidity.
Genetics + Functional Genomics
Variants, ClinVar, GWAS, expression, sequencing, perturbations, single-cell studies.
Medication + PGx
RxNorm, labels, mechanisms, targets, pharmacogenomics, safety signals.
Environment + Epidemiology
Exposure, toxicology, pollution, population burden, geography, demographics.
Similar Cases
Phenotype, labs, genetics, medications, exposures, course, outcomes—without claiming identical disease.
Evidence Board + Notebook
Pin source-backed findings, organize questions, reproduce analysis in Jupyter.
AI Evidence Brief + Clinician Questions
Grounded synthesis with citations, uncertainty, contradictions, and discussion questions.
Go back to the original scientific source
MedlinePlusPlain-language health and disease information from the U.S. National Library of Medicine.
PubMedBiomedical literature citations and abstracts.
Europe PMCLiterature, abstracts, full-text links, grants, and biomedical entities.
ClinicalTrials.govRegistered clinical studies and trial results.
ClinVarVariant interpretations with review status and disease context.
Open TargetsDisease-target associations, genetics, drugs, and evidence integration.
Human Protein AtlasProtein expression across tissues, cells, and disease contexts.
RCSB PDBExperimental 3D structures of proteins, DNA, RNA, and complexes.
openFDALabels, recalls, adverse-event and regulatory datasets.
EPA CompToxChemical, toxicology, bioactivity, and exposure-related research data.
CDC WONDERU.S. public-health statistics with dataset-specific use restrictions.
WHO Global Health ObservatoryGlobal health indicators and disease-burden data.