Short answer. DNA is excellent at identity (same plant or not), exclusion (this can't be the parent) and trait markers with proven links (sex, chemotype class, some flowering genes). It is weaker at inclusion without both parents sampled, and useless for proving an old story attached to a name. It does not measure potency, flavour or effect.
What's happening
A genotype test reads many DNA positions (SNPs or microsatellites) in a sample. Comparing those readings across samples answers questions about relatedness. Parentage logic is simple [1]: a child must carry one allele from each parent at every position. One consistent mismatch across many markers (allowing for error) excludes a parent; consistent matches support but don't prove parentage, because relatives of the true parent may also match.
Can and can't
| Question | Can a test answer it? | Notes |
|---|---|---|
| Are these two clones the same genotype? | Yes, with enough markers | Expect tiny differences from somatic mutation and genotyping error [7] |
| Is plant X a possible parent of seed lot Y? | Yes, by exclusion | Strongest when several offspring are tested |
| Is plant X the parent? | Supported if both parents and offspring are sampled and close relatives are ruled out | This is the L4 standard |
| Is this the "original" 1990s cut? | Only by comparison with a documented reference sample | Without a reference, no |
| Will this seedling be male? | Yes with validated sex markers [3] | |
| Is this seedling THC- or CBD-dominant? | Usually, chemotype class markers [3] | Markers can fail in some germplasm |
| Is this an autoflower? | Sometimes [6] | Markers don't predict all auto sources |
| What THC % will it test? | No | Total cannabinoids are quantitative and environment- and lab-dependent |
| What does it taste like / feel like? | No | Terpene synthase variants are linked to labels [5], but aroma is multi-gene and environmental |
| What % indica vs sativa is it? | Not meaningfully | Genome-wide, "indica" and "sativa" labelled samples were genetically indistinct in [5]; such percentages describe similarity to a reference panel, not ancestry |
| Which landrace is it from? | Only roughly, if the reference panel contains relevant landraces | Reference panels are biased to commercial material [2] |
What to ask a lab
- What markers and how many? Thousands of SNPs give far better resolution than a few microsatellites. Schwabe and McGlaughlin used 10 microsatellites [4]; Henry et al. used 22 targeted SNPs chosen for ancestry and chemotype across 420 accessions [2]. Both are useful research panels; neither is a whole-genome fingerprint.
- What reference database? Matches are only as good as the samples behind them.
- Error rate and repeatability? Ask for replicate data.
- How do they report relatedness? Look for identity-by-state or kinship values, not marketing labels.
- Can they run parentage with your own parents and offspring? This is what L4 needs.
- Data ownership: can you download raw genotypes, and who else can use them?
Step by step: an L4 parentage test
- Sample the claimed female, the claimed male (or reversed female) and at least 5 offspring. Use young leaf tissue, labelled with GB-IDs.
- Record sampling with dated photos showing the label and plant.
- Send to the lab with a chain-of-custody note.
- Request a parentage or kinship analysis.
- Attach the report as a GB-DOC to the LIN claim. A staff archivist checks consistency.
Common mistakes
- Taking "98% match to Strain X" as identity. Ask: match to what sample, from where?
- Treating "62% sativa" as ancestry.
- Assuming a DNA match means the same flower quality. Health (HLVd), environment and somatic changes still matter.
Evidence
- Parentage logic and exclusion vs inclusion: Established [1].
- Clone identity and population structure from SNP assays: Supported [2][4].
- Sex and chemotype markers: Supported [3].
- Indica/sativa labels not reflecting genome-wide ancestry: Supported [5].
- Autoflower markers: Supported in mapped germplasm [6].
