Genetics, seeds and the pheno huntGenetics, seeds and the pheno hunt

Purpose and scope

Every cultivar worth growing started life as one individual plant that somebody noticed, kept and copied. This paper explains the machinery behind that: what genetics can and cannot promise, why two seeds from the same pack grow into different plants, what the words on a seed listing actually mean, and how to run a phenotype hunt, a structured search through seeds for the one plant worth keeping.

It is written for someone starting from zero, but the process at the end is the same one commercial operators use, just at different scale. No prior genetics knowledge is assumed; every term is defined the first time it appears.

The paper runs in grow order: the three layers of what a plant is, where cultivars came from, why seeds vary, the seed types you can buy, the hunt itself, selection and sample size, then keeping, testing and breeding from the winner.

Evidence and limitations

Genotype, phenotype and chemotype

Three words carry the whole subject. Get them straight and everything downstream (seed types, hunts, testing, breeding) becomes simple mechanics.

The layers are linked but not interchangeable. The THC:CBD ratio is close to hard-wired: it is controlled mainly by one genetic locus with two versions (alleles), and crossing a pure THC plant with a pure CBD plant gives offspring that split into three chemotypes in a predictable 1:2:1 pattern.^[1] The absolute potency and terpene numbers, though, move with the grow, light, ripeness, health, because they sit in the environment-facing phenotype layer.

Landraces, polyhybrids and unreliable strain names

Cannabis was domesticated in East Asia around the early Neolithic, and everything grown today, hemp and drug types alike, descends from that ancestral pool.^[3] As people carried it around the world, isolated regions developed landraces: locally adapted, open-pollinated populations shaped by their climate and their farmers over many generations.

Modern drug cannabis is overwhelmingly polyhybrid. Genome-wide studies show the familiar labels only loosely track reality: the reported ‘sativa’ or ‘indica’ ancestry of commercial strains corresponds only moderately to their actual genetic structure.^[4] Worse for the shopper, samples sold under the same strain name from different sources are frequently different plants: microsatellite fingerprinting of dispensary samples found genetic inconsistency within most strain names tested.^[5]

None of this means genetics do not matter. They matter enormously. It means the name is a weak label for them. A name buys you a rough style guide (probable aroma family, rough structure) and nothing bankable.

Seed variation: heterozygosity and segregation

Cannabis is naturally an outcrossing species, separate male and female plants, wind pollination, constant mixing. That history makes it highly heterozygous: at many positions in the genome, each plant carries two different versions of the gene.

This is why seed-grown plants differ: every seed is a new random draw from both parents' decks. If both parents are true-breeding (homozygous) for the traits you care about, the first generation, the F1, is uniform, because every seed draws the same hand. Cross that F1 with itself and the F2 explodes into variety as the alleles recombine. The classic demonstration is chemotype: pure-CBD × pure-THC parents give an all-mixed F1, and the F2 splits 1:2:1 into CBD-dominant, mixed and THC-dominant plants.^[1]

Maize breeders solved this a century ago with inbred parent lines and true F1 hybrid seed. Cannabis mostly has not: prohibition kept breeding informal, so the industry runs on heterozygous parents and the variation lands in your tray. A pack of ten seeds is ten related individuals, siblings, not copies.

Genetic stability

Seed catalogues throw around F1, IBL and ‘stable’ loosely. Here is the vocabulary with its real meaning, so you can read a listing critically.

Inbreeding is a trade. Each generation of selfing or sibling crossing roughly halves the remaining heterozygosity, which stabilises traits. But cannabis is an outcrosser: repeated close-crossing allows harmful recessive alleles to pair up and get expressed, sapping the plant's health and productivity. That cost is called inbreeding depression. The long-term prize is the maize model: two inbred parents crossed to make true F1 seed that is both uniform and vigorous. A handful of seed companies are now working exactly that way; most of the market is not there yet.

Seed types: regular, feminised, autoflower, S1 and clone-only

Every seed on the market is one of a few constructions, and the construction tells you the sex ratio, the variation to expect, and what the seed is for.

Regular seed is the natural cross: male pollen onto a female plant. Roughly half the seedlings will be male, which flower growers cull, males make pollen, not bud. Regular seed is the cheapest per seed, carries the full genetic shuffle, and is what breeding programmes need, because it is the only type that yields males.

How feminised seed is made. And why it is not ‘weaker’

Feminised seed comes from pollinating a female with pollen from another female that has been chemically persuaded to grow male flowers. Cannabis plants use a hormone called ethylene as a signal that keeps them in female mode — think of it as a thermostat set to keep the plant female. Block that signal and male flowers develop, even on a plant that is genetically female. Silver thiosulfate (STS) does the blocking: silver ions bind to the plant's ethylene receptors and silence the signal.^[6] In practice a dilute STS solution is sprayed on a mother a few times around the flip, and she produces pollen a few weeks later.^[7] Because that pollen comes from a plant with two X chromosomes, every seed it makes is XX, female. Properly made feminised seed runs at or near 100% female in published trials.^[8][9]

The ‘fem seeds are weak / hermie-prone’ folklore confuses the method with the parents. STS changes hormone signalling on the mother for a few weeks; it does not mutate the DNA that goes into the seed. Where feminised seed earns a bad reputation is parent choice: seed made by stressing plants until they self-pollinate (rodelization) actively selects for the tendency to throw male flowers under stress, and that tendency is heritable.^[10] Ask how the seed was made. STS reversal of a stable, tested mother is the standard; stress-derived seed is the lottery.

Autoflower seed carries a day-neutral flowering trait inherited from Cannabis ruderalis: the plant flowers on age, not photoperiod. The major locus behind it (Autoflower1) behaves as a simple recessive, which is why both parents must carry it and why crossing an auto to a photoperiod plant gives photoperiod offspring that merely carry the allele.^[11] Autos trade ultimate size and the ability to hold a mother plant (you cannot keep a plant in veg that flowers on age) for speed and simplicity.

S1 seed is a plant selfed: a female reversed with STS and used to pollinate herself. The offspring cluster around the mother's character but they are not copies. Every locus where she was heterozygous still segregates. An S1 of a famous cut is a neighbourhood around the cut, not the cut.

Clone-only means the cultivar exists only as a vegetatively copied cut, there is no seed line. That is what a keeper becomes after a hunt, and it is the only way to hold one exact genotype over time. See the cloning paper for the mechanics.

Phenotype-hunting workflow

A pheno hunt is a controlled comparison: grow a batch of seeds under conditions as identical as you can manage, score every plant against criteria fixed in advance, and keep the best individual as a clone. The enemy is confounding, any difference in position, pot, feed or timing that lets a mediocre plant look special, or hides a great one.

1. 1
   Size the hunt before you germinate

   Decide how many seeds your space can carry to harvest as one cohort, and be honest about the odds that number buys (next section). Write the scoring rubric now, before you have favourites.
2. 2
   Pop everything at once, tag everything

   Germinate the whole batch together (see seeds and germination). At first true leaves, give every plant a permanent ID, pack code plus seed number, and make the tag follow the plant through every transplant. An unreadable hunt is a wasted hunt.
3. 3
   Sex early and cull males (regular seed)

   Photoperiod plants show sex at preflowers around week 4–6 of veg; a leaf-swab genetic test can call sex weeks earlier at the seedling stage.^[12] Unless you are breeding, males leave the room before any pollen sac opens.
4. 4
   Veg to a fair comparison

   Same pots, same medium, same feed, same topping policy for every plant. Note veg vigour and rooting speed in the log, but resist selecting on them. The call happens after flower, on the full picture.
5. 5
   Take backup cuts of every candidate

   Two or three cuttings per plant, labelled with the parent's ID, rooted and parked in veg before the flip. This is the step beginners skip and regret: flower reveals the winner, and without a cut the winner is already dead when you meet it.
6. 6
   Flower identical, rotate positions

   One room, one recipe, all plants flipped together. Rotate positions weekly so edge effects and hot spots average out instead of crowning whoever stood under the best light.
7. 7
   Score weekly against the rubric

   Stretch, structure, onset of flowering, pest and mould events, aroma as it develops. Write numbers, not vibes. Do not crown anyone at week 3 — loud early terps are not a finished plant.
8. 8
   Harvest, weigh and assess per plant

   Dry weight per plant, kept separate through dry and cure. Send samples for testing if you can; if hash is the goal, run a small wash or press trial per candidate, flower quality and resin yield are different traits.
9. 9
   Verify the finalists in round 2

   Flower the backup cuts of your top two or three side by side. The keeper is the one that repeats its performance as a clone. One good run is an audition; two is a cultivar.

Selection criteria beyond potency

Potency is the loudest criterion and the worst one to select on alone. A 26% plant that moulds every autumn, stretches into the lights and roots badly is a liability with a good lab number. Operators score across the whole job the plant has to do:

Sample-size limitations

Here is the arithmetic nobody puts on the seed pack. Suppose a genuine keeper, a plant that clears your bar on every criterion, shows up in about one seed in twenty from a decent cross. That 5% is generous for a strict rubric, and it compounds like this:

This is not an argument against small hunts. It is an argument for honest language. Ten seeds reliably finds the best plant you had, and that plant may well be worth keeping and growing for years. It is just unlikely to be the once-in-a-line individual that commercial hunts chase by popping hundreds to thousands of seeds and keeping one or two. Selection intensity is the whole difference: best-of-10 and best-of-500 are different animals wearing the same word.

Maintaining a keeper cut

The hunt ends with the most valuable object in the facility: one plant. From here the job is redundancy. A keeper held as a single mother is one root-rot event, one viroid infection or one labelling mistake away from not existing.

Screen before you commit. Hop latent viroid (HpLVd), the ‘dudding’ pathogen, spreads silently through cuttings and tools, and infected stock can look normal for months; test the candidate before it becomes a mother, and treat one negative as provisional rather than proof, because low, uneven viroid levels can slip past a single test.^[13]

For long-term insurance, a tissue-culture archive holds the genotype in clean storage off the grow floor. Archive early and at low passage: plants held in culture accumulate small somatic mutations roughly in proportion to how many times they are subcultured, so the best archival copy is made once, young, and disturbed as little as possible.^[14]

Genetic testing capabilities and limitations

Cheap genetic assays now cover three jobs that used to cost weeks of grow time. All three run off a small leaf sample.

The boundary to hold in your head: a swab reads the genotype layer. Sex and chemotype class live there, so markers call them well, the synthase-gene region they probe is well mapped, if messy.^[2] Potency numbers, terpene character, vigour and yield live in the phenotype layer, shaped by the grow. No swab predicts them, whatever the marketing says.

Breeding basics for cultivators

Once you hold a keeper, the next itch is making seed from it. Two modes exist. Open pollination, males and females loose in one space, is how landraces work: maximum recombination, zero control, fine for making a big diverse seed batch from a population you like. A controlled cross is one chosen father onto chosen branches of one chosen mother, and it is the only way to know what you made.

1. Isolate the male. A separate space with separate airflow, shared HVAC is shared pollen. Let it open its first flowers over paper or glass.
2. Collect and dry the pollen. Tap it free, let it dry for a day or two, then pass it through a fine sieve to remove flower debris.
3. Store it cold and dry. Small airtight vials with a desiccant, labelled, in the freezer. Viability fades over months, use fresh where you can, and test a pinch on one branch before trusting a stored batch.
4. Pollinate selectively. Paint pollen onto a few lower branches of the mother with a small brush, tag those branches, and mist nearby surfaces afterwards, water kills stray pollen.
5. Wait, then harvest seed. Seeds mature in roughly 4–6 weeks; ripe ones are dark, hard and striped. Dry them with the flower, then store cool, dark and dry.

The reason for all the ceremony: pollen is nearly invisible and absurdly effective. One open male, or one stress-induced hermaphrodite, can seed a whole flower room, and seed set by accidental self-pollination quietly carries the parent's instability forward.^[10] Breeding in the same building as sinsemilla production is a containment exercise first and a romance second.

Intellectual property and licensing

Cultivar ownership is real but patchy, and it varies by jurisdiction. A few generic truths hold. Strain names are mostly unprotected marketing, and as covered earlier, often do not even track a consistent genotype.^[5] Actual protection, where it exists, attaches to the plant material or the registered variety: plant variety rights / plant breeders' rights schemes, patents in some countries, and increasingly, contract terms attached to licensed clone releases, nurseries supplying verified cuts under agreements that limit propagation, resale or breeding.

Practical hygiene for an operator, anywhere: keep records of where every cultivar came from and under what terms; read the terms on licensed cuts before breeding from or distributing them; and treat your own keeper's provenance log, hunt records, dates, test results, as the documentation you would want if you ever release or license it. For anything beyond that, the rules are local: check them where you are before selling genetics in any form. This is orientation, not legal advice.

Common phenotype-hunting failures

Most failed hunts fail the same six ways, and every one is preventable for the cost of discipline.

Troubleshooting

Phenotype hunting: selection, validation and preservation

From here, the practical neighbours: seeds and germination for getting the tickets sprouted, cloning for taking and rooting the cuts the hunt depends on, and tissue culture for the archive that makes a keeper permanent. The genetics do not care what the packet said: pop enough seeds, score them honestly, verify the winner, and protect the cut with the same discipline you brought to finding it.