Indica or sativa? The DNA barely agrees with the label

The "sativa" that put you on the couch#

Almost everyone has a story. The "energising sativa" that glued you to the sofa. The "heavy indica" that had you reorganising the kitchen at midnight, then alphabetising the spice rack. Budtenders explain it with tolerance, terpenes or the time of day. We wanted to know whether the label itself means much.

So we tested it. CannaLib holds DNA genotypes for 3,584 samples from four public datasets, each matched to the strain name it was sold under, and stated types for 72,945 strains. For every sample with a stated type, we asked one question: can the measured genetics predict the label?

How we tested it#

We turned each label into a number (indica 1, indica-dominant 0.75, hybrid 0.5, sativa-dominant 0.25, sativa 0) and asked whether the biggest patterns in each panel's DNA (the stuff that sets plants apart most) could predict it for samples the model hadn't seen. We also ran the same test 500 times with the labels shuffled at random, as a "what if it's all luck?" control.

The score is called R², and here's the plain-English version. Imagine guessing a jar's label from its DNA alone. A score of 0% means you may as well flip a coin, which is exactly what the shuffled labels scored. A score of 100% means the DNA gives the label away every single time. Anything in between is "how much of the mystery the DNA solves". We ran it on the stated type and on heritage percentages like "70% indica".

The label isn't random noise. In every panel it beats chance. But in the panel that looks most like a modern dispensary menu, 1,048 samples sold under 618 names, the DNA predicts only about 15% of the label. That's like watching a movie trailer and trying to guess the ending: you get the vibe, not the plot. Heritage percentages do worse, at 9%.

The label works better in the Ferguson panel, at 46%, and the reason matters. That panel mixes hemp, feral plants and drug strains, populations that really are far apart. So the label is decent at telling broad groups apart, the way anyone can tell a cat from a dog. It's poor at telling one modern hybrid from another, which is like telling two tabbies apart, and that's exactly the choice you're making at the counter.

The label is good at telling hemp from a dispensary strain. It's poor at telling one dispensary strain from another.

This fits what researchers have been finding for a decade. A 2015 study of 81 drug strains concluded that reported indica and sativa ancestry often failed to match the genetics [2]. A 2021 study found labels tracked differences at a few aroma genes rather than overall ancestry [3]. In other words, "indica" may describe how a strain smells more than where it came from.

What the label rests on#

We also looked at the evidence behind the type field in CannaLib itself. Of 72,945 strains with a stated type, most rest on a seller's or a website's say-so. Only 1,020 (about 1.4%) have that type supported by independent evidence such as inheritance, published grower literature or terpene profiles. Another 704 are disputed, where sources or evidence disagree. The rest is a lot of "trust me, bro" on a lot of labels.

What the DNA does see#

Here's the twist. The DNA isn't blind to how cannabis behaves. It just isn't tracking the indica / sativa word. The thing it tracks clearly is the balance of THC and CBD. Plot twist: the small print beats the big print.

That matches what lab tests show. CannaLib has measured the chemotype of more than 8,000 strains from two or more agreeing lab tests. 7,761 (about 96%) are THC-dominant, 219 are CBD-dominant and 77 are balanced. If you want a hint at how a strain will feel, its measured THC and CBD are a better starting point than the word indica or sativa.

So what should you look at?#

Check the measured THC and CBD range, ideally from several lab tests. Look at the terpenes the labs actually found. And trust your own notes. Your body's response to a specific strain is data too, and it's the data we have least of. You're basically a lab of one, with snacks.