The tube is not the test
You spat, mailed a kit, and received a story. Or a clinician ordered blood and a laboratory sent a PDF. Both involve DNA. They are not interchangeable. Types of DNA tests differ in what they look at and what they are allowed to claim.
A genotyping chip asks a predetermined list of common SNPs which of two letters is present. Sequencing reads letters in order along a stretch of DNA. Those are different machines. A chip cannot invent a rare variant it was never designed to see. A sequence of one gene cannot describe your ancestry in 25 regions.
MedlinePlus splits clinical tests by scope: a single known variant, a whole gene, a panel of genes, an exome, a genome. Direct-to-consumer products sit to the side of that list. They are sold to you, not ordered as a diagnostic assay, and extra testing is usually required before anyone should treat the result as care.
Ancestry chips: hundreds of thousands of bookmarks
An ancestry DNA test is almost always a genotyping chip. It samples on the order of 600,000 common SNPs across the autosomes, sometimes with a thin extra layer of Y or mitochondrial markers. The company then compares your pattern with a reference panel and prints ethnicity estimates, relative matches, and perhaps a haplogroup.
What it can see: common variation useful for matching people to people and to reference groups. What it cannot see: most rare disease alleles, most of BRCA1, structural rearrangements, and the other 3 billion bases the chip never touched.
The ethnicity bar chart is a model, not a passport. Change the reference panel and the percentages move. That is what a model does. Treat it as a statistical resemblance, then read the health-versus-ancestry split separately if the same company also sold you a wellness report.
- Sees: a designed list of common SNPs, plus relative matching
- Does not see: most of any gene, most rare pathogenic variants
- Output: ethnicity estimates, matches, sometimes a haplogroup
- Not: a medical diagnosis, a whole genome, a BRCA test
Health-report chips: the same technology, a different brochure
A health-report chip is often the same spit, the same array, and a second layer of software. The company highlights a subset of SNPs with FDA-authorized or in-house reports: carrier status for a few variants, a pharmacogenetic tag, a well known risk allele. The raw-data file underneath is still a table of rsIDs and two letters.
What it can see is whatever probes were on the chip and whichever of those the company chose to interpret. What it cannot see is the rest of each gene. A report that says you do not have a particular founder variant is a statement about that variant. It is not a negative test of the gene.
Direct-to-consumer health reports are not clinical care. MedlinePlus is explicit: additional testing through a healthcare provider is usually required before those results are used to diagnose a condition or make care decisions. A red badge in an app is not a referral.
Clinical single-gene tests look hard at one place
When a condition has a short list of usual genes, a clinician may order a single-gene test. The laboratory sequences that gene, and often looks for deletions or duplications that sequencing of short fragments can miss. The question is focused: is there a disease-causing variant in this gene that fits this person's presentation, or this family's known variant?
A targeted single-variant test is even narrower. It looks for one known family spelling, the way a relative of someone with a documented HBB change might be tested for that change. Direct-to-consumer companies also use this narrow style when they report a handful of famous sites rather than the whole gene.
What a single-gene test cannot do is rule out every other gene. People with the same symptoms can have different genetic causes. If the first gene is clean, the clinical question is not over. It has only finished one chapter.
Panels cast a wider net on purpose
A gene panel sequences many genes at once because the phenotype does not point at a single suspect. Epilepsy, cardiomyopathy, and cancer-predisposition panels exist for this reason. Hundreds of genes can cause overlapping stories, so testing them together can be faster than a queue of single-gene assays.
A panel is still a list. Variants in genes that were not on it remain invisible. Coverage inside the list is not always perfect. Some exons are hard to sequence. Some copy-number changes or repeat expansions need extra methods.
Panels are clinical tools. They come with a reason for testing, a laboratory's validation, and usually a path to genetic counseling. They are not a more thorough 23andMe. They are a different class of measurement.
- Single variant: one known spelling, often a family test
- Single gene: one locus, sequenced and often copy-number checked
- Panel: many genes chosen for a phenotype
- Chip: common SNPs, no claim to complete any gene
Exomes read the protein-coding 1 percent
The exome is the collection of exons, the DNA that is translated into protein. It is about 1 percent of the genome and contains most of the variants that are currently known to cause Mendelian disease. Whole-exome sequencing tries to read that 1 percent in one assay.
About 20,000 protein-coding genes live in that space. An exome can find a rare missense or a frameshift that no chip would have probed. It can still miss variants that sit deep in introns, in regulatory DNA, or in regions the capture kit did not cover well. It is a powerful clinical tool with a known ceiling.
An exome is also a flood of uncertain findings. Most of what it sees has no medical meaning. Secondary findings, variants in genes that were not the reason for the test, are a separate counseling problem. More sequence is not automatically more answers.
Whole-genome sequencing tries to read the book
Whole-genome sequencing, WGS, determines letters across nearly all three billion bases, not only the exons. In principle it can see SNPs, many indels, some structural changes, and variation outside genes. In practice, some regions remain hard, and the meaning of most noncoding changes is still unknown.
WGS is not a consumer chip with better marketing. It is a different assay. It still is not a crystal ball. A genome can be sequenced and remain unsolved if the variant is missed, misinterpreted, or not actually genetic in the way the question assumed.
Cost and interpretation, not saliva, are the usual limits. A genome file is identifying in a way a chip file already was, only more so. Privacy does not get easier when the assay gets larger.
Carrier screening asks a family-planning question
Carrier screening looks for variants that usually do not make the carrier ill, but that can cause recessive disease in a child if both parents pass a pathogenic variant in the same gene. Cystic fibrosis, spinal muscular atrophy, and hemoglobinopathies are classic examples. The point is reproductive risk, not a diagnosis of the person being screened.
Clinical carrier screening can be a short ethnicity-based list or a large pan-ethnic panel. Laboratories validate the genes and the variant types they claim to see. Some conditions need special assays, such as the SMN1 copy-number test, that a SNP chip does not replace.
A health-report chip that includes a few carrier variants is a partial overlap, not the same product. A negative chip report for one founder allele is not a negative carrier screen for the gene. If the question is family planning, the assay has to match the question.
A 23andMe file is not a BRCA test
BRCA1 and BRCA2 are large DNA-repair genes. Pathogenic variants in them raise the lifetime risk of some cancers. Clinical BRCA testing sequences the genes and looks for copy-number changes. Founder variants exist in some populations and are famous because they are famous, not because they are the whole gene.
A consumer file contains a thin sample of common SNPs. A company may report a handful of BRCA sites. A normal result at those sites is not a negative BRCA result. It is a sampling gap. Most pathogenic BRCA variants will never be on the array.
The same warning applies across genes. HFE, HBB, and APOE show up in tutorials because a few SNPs are well known. They are useful literacy, not a substitute for the matching clinical test. Read a file you already have as a table of called SNPs, then look up the public record without pretending the chip saw everything.
Questions
What are the types of DNA tests?
Common types include ancestry chips, health-report chips, clinical single-gene tests, gene panels, exome sequencing, whole-genome sequencing, and carrier screening. They differ in what they can see, not only in how the sample was collected.
Is a 23andMe test a whole genome?
No. A 23andMe file is a genotyping chip: hundreds of thousands of predetermined SNPs, not three billion sequenced bases. Consumer chip is not whole-genome sequencing.
What is the difference between an exome and a genome test?
An exome reads protein-coding exons, about 1 percent of DNA and about 20,000 genes. A whole genome attempts to read nearly all of the letters, including DNA between genes.
Is a 23andMe file a BRCA test?
No. Clinical BRCA testing sequences BRCA1 and BRCA2 and looks for copy-number changes. A chip that reports a few founder sites has not tested the genes.
What is carrier screening?
A test for variants that usually do not make the carrier ill but can cause recessive disease in a child if both parents pass a pathogenic variant in the same gene. A few chip reports are not a full carrier panel.
Can an ancestry DNA test diagnose a disease?
No. Ancestry chips are built for common SNPs and relative matching. Diagnosis needs a clinician and an assay matched to the condition. This page is not medical advice.
Why did my health DNA report miss a variant my clinician found?
Because the chip never looked there, or looked only at a different site in the same gene. Missing from a consumer export means not called, not absent from your genome.
Which DNA test is the most complete?
Whole-genome sequencing sees the most letters, with remaining gaps. Completeness is not the same as usefulness. The right test is the one that can answer the clinical or ancestry question you actually have.