Glossary

Plain-language definitions for the terms that come up across the atlas. Each entry is one or two sentences.

Alarmin: A signal cells release when they are damaged or stressed, which starts inflammation to defend the area. Helpful briefly, harmful when it will not switch off. IL33 and TSLP are alarmins.
Allele: One of the different versions of a gene or DNA spot. You inherit one copy from each parent, and they can match or differ.
Ancestry / cross-ancestry: Your genetic ancestry is the population history carried in your DNA. "Cross-ancestry" asks whether a finding from one population (often European) actually holds in others.
cis-eQTL: A DNA variant that changes how strongly a nearby gene is switched on. "Cis" means it acts on a gene close to it on the same chromosome.
Cohort: A group of people studied together, often over time. Larger, more diverse cohorts give more trustworthy results.
Confidence rating (low / medium / high): GENARCH's own label for how solid a claim is, based on how many independent lines of evidence agree. High means several methods point the same way; low means a single hint worth flagging but not settled.
Confounding: When a hidden third factor makes two things look related even though one does not cause the other. Careful studies try to rule it out.
Cytokine: A small protein cells use to talk to each other, mostly to run the immune system. Some cytokines ramp inflammation up, others calm it down.
Effect size: How much influence something actually has, not just whether it is real. A variant can be statistically certain yet shift risk only slightly.
eQTL: A DNA variant linked to how strongly a gene is expressed. It helps connect a risk variant to the gene it actually affects.
Epidemiology: The study of how often diseases occur across populations and why. It looks at group patterns, not individual diagnosis.
Exposure: Something in the environment that can affect health, such as air pollution, diet, or limited food access. GENARCH focuses on exposures you could plausibly change.
Feature importance: In a model, a ranking of which inputs matter most to the prediction. See SHAP.
Gene expression: How much a gene is actually switched on in a given cell or tissue. The same gene can be loud in the lung and quiet in the liver.
Gene–environment interaction (GxE): When genes and surroundings combine so the effect of one depends on the other. The same pollution can matter more for some genotypes than others.
Genome-wide significance: A strict statistical bar (p < 5×10⁻⁸) a genetic association must clear to count, set high because so many variants are tested at once.
GTEx: A public reference dataset showing which genes are expressed in which human tissues. GENARCH uses it to add tissue context.
GWAS (genome-wide association study): A scan of DNA across many people to find variants more common in those with a disease. It points to regions of interest, not exact causes.
Heritability: The share of the differences in a trait across a population that ties to genetic differences. It is a population number, not a percentage for any one person.
Incidence: How many new cases of a condition appear in a population over a set time. Prevalence counts existing cases; incidence counts new ones.
Locus (plural loci): A specific location on the genome. In GWAS it usually means a region flagged as linked to a trait.
NF-κB: A control switch inside cells that turns on inflammation genes in response to stress or pollution. It shows up as a hub in several disease mechanisms here.
Odds ratio: A number comparing the odds of disease with a variant versus without. Above 1 means higher odds, below 1 means lower.
Pathway: A chain of genes and molecules that work together to get something done in a cell. Disease often comes from a pathway being pushed too far.
Penetrance: How reliably a variant actually leads to the trait. High penetrance means most carriers show it; low means many do not.
PM2.5: Fine airborne particles under 2.5 microns, small enough to reach deep into the lungs. Traffic and industry are common local sources.
Polygenic: Shaped by many genes at once, each adding a little, rather than one gene deciding the outcome. Most common diseases are polygenic.
Prevalence: The share of a population that has a condition at a given time. It is the main way GENARCH describes community health burden.
PRS (polygenic risk score): A single number that sums the small effects of many variants to estimate genetic predisposition across a population. GENARCH does not compute personal scores, and these scores lose accuracy across ancestries.
Replication: Getting the same result again in a separate, independent study. A finding that replicates is far more trustworthy than one that does not.
SHAP: A method that explains a model's prediction by showing how much each input pushed the result up or down. Positive values raise the predicted burden; negative values lower it.
SNP / rsID: A SNP is a single-letter difference in DNA between people. Its rsID (like rs7216389) is the standard catalog name for that spot.
Tissue specificity: The fact that a gene or mechanism can matter in one tissue but not another. It is why GENARCH ties genes to specific tissues like bronchial epithelium.
Variant: Any place where one person's DNA differs from another's. Some variants do nothing; a few shift disease risk.