Coffee genetics and breeding: from genome to variety

Two genomes in one plant

Sourced

Most coffee species carry 22 chromosomes. Arabica carries 44. A genome study published in Nature Genetics in 2024 describes it as an allotetraploid: the product of a natural cross between the ancestors of present-day Coffea canephora, the species farmed as robusta, and Coffea eugenioides, with each parent's full set retained. Geneticists write the two halves as subgenomes. The same study reports that the robusta-derived half is most closely related to canephora from northern Uganda.

The two halves do not usually mix. The study describes arabica as generally pairing each chromosome with its own partner and inheriting its genes as an ordinary plant with two sets would, with occasional exchanges between the halves. One such exchange, at the end of chromosome 7, is shared by every plant its authors examined, wild and cultivated, and they infer that it dates from soon after the founding cross. An earlier study, which sequenced a Bourbon tree, had found the same replacement.

When the cross happened is a matter of inference, and the two studies do not agree. The 2024 authors, modelling population history from 41 arabica genomes, put the founding event between about 610,000 and 350,000 years ago, and say that much more recent estimates in earlier work may be underestimates caused by later bottlenecks. The 2020 study, in Scientific Reports, describes the bottleneck as recent and severe. These are modelled dates. They depend on an assumed generation time, 21 years in the 2024 paper, and on a mutation rate, and neither is a measurement.

Why arabica trees are so alike

Sourced

The 2020 study genotyped 736 arabica plants: cultivated varieties, the large collection held in Costa Rica, and 93 trees from farmers' fields in Yemen. Its authors report the lowest genetic diversity yet recorded in a crop species, comparable only to bread wheat, and attribute it to a single founding event affecting the whole species and not only what is farmed. They draw a blunt conclusion for breeding: diversity inside arabica is less useful than in most crops, and new variation has to be brought in from the two parent species.

The 2024 study agrees on the fact and adds a history. It found that wild arabica was already low in diversity before anyone cultivated it, and that cultivated plants are only slightly less diverse than wild ones. That is unusual. Domestication normally starts from a varied wild plant and narrows it; here there was little to narrow. Its models suggest that the lineage leading to cultivated coffee separated from the other wild populations about 30,000 years ago and went on exchanging genes with them until roughly 8,000 to 9,000 years ago.

On top of that came the bottlenecks of trade, which are history and not inference. The same paper summarises them: cultivation in Yemen from the fifteenth or sixteenth century; a plant shipped to Amsterdam in 1706 from which Caribbean cultivation was founded in 1723; and on the island of Bourbon, now Réunion, the descendants of a single plant that survived to 1720. Its kinship analysis found Typica, Bourbon and an old Indian selection related as parent and offspring, which it takes as confirmation of a shared Yemeni origin and of how narrow that Yemeni stock was. Bourbon was less diverse than Typica, as a group founded on one plant would be.

What Ethiopia still holds

Sourced

Low diversity is not no diversity. The 2020 study found a gradient running west to east, from the forests of south-western Ethiopia to the plantings of eastern Ethiopia and Yemen. The 2024 study sampled wild trees on both sides of the Great Rift Valley and found that, of the seventeen wild trees in its sample, one of the two genetically closest to the presumed wild parent of cultivated coffee had been collected near Gesha mountain in the west. Its authors call the Gesha region a hotspot of wild trees amenable to domestication. They could not identify where the cultivated population originated.

A resequencing study of 90 trees from an Ethiopian field gene bank, covering both garden and forest coffee, sorted them into four clusters on a family tree built from their DNA. Forest trees from the south-west turned up in several clusters, which its authors read as a broad genetic base there; trees from the south-east mostly fell together; and garden coffees collected in the same area sat beside one another. The paper reports a count of variants and that tree. It gives no measure of diversity that could be set beside the 2020 study's, and its introduction notes that Ethiopia's field gene banks hold more than 11,000 accessions, of which 90 is a small sample. Together these studies explain why breeders value Ethiopian material out of proportion to its measured diversity: it is the part of the species that the Yemeni bottleneck did not pass through.

A study led from Kew, described in the atlas entry for Ethiopia, is the source for how exposed those wild populations are. The trade's word for Ethiopian coffee of unrecorded parentage, heirloom, names that gap in the records and not a variety.

The Timor Hybrid: a second cross, a century ago

Sourced

Because the two halves of arabica's genome are so close to their parent species, arabica can still occasionally cross with robusta. One such spontaneous hybrid, resistant to leaf rust, was identified on the island of Timor in 1927. Crossed back to arabica varieties over the following decades, it is the source of the rust resistance in the Catimor and Sarchimor families and in many varieties bred since.

The 2024 genome study sequenced the Timor Hybrid and five of its descendants and measured what the cross left behind. The robusta material sits almost entirely in the robusta-derived half of the genome, in large unbroken blocks, as expected of something only a century old, and covers 7 to 11 per cent of the genome in the lines examined. One region on chromosome 4 is shared by every descendant. It contains three clusters of genes of kinds known from other plants to be involved in disease resistance, and the authors name eight of them as strong candidates for rust resistance. Candidate is their word: the paper locates the region and does not show which gene does the work.

Introgression is the term for genetic material carried from one species into another by crossing and repeated backcrossing. It brings neighbours with it. A block of chromosome arrives whole, with whatever other genes it holds. The genome paper's introduction says that the Timor-derived introgressions ensure rust resistance but also have unwanted side effects such as lower beverage quality. That is a statement of the field's general view, placed in an introduction, and not something that study tested. The variety pages on this site give the evidence variety by variety, and it is more mixed than the folklore.

One species fertilises itself, the other cannot

Sourced

Arabica is self-fertile. A tree pollinated by itself sets seed, and after several generations of this its offspring are close to identical with it. For more than a century arabica breeding therefore meant selecting good trees and fixing them as pure lines, which come true from seed. A review of hybrid breeding in arabica, written in 2026 by three breeders, describes this as the strategy most national programmes adopted, and gives one of the reasons: seed of a pure line is cheap to produce.

Robusta is the opposite. A review of a century of robusta breeding, published in 2025, describes the species as heterozygous and predominantly outcrossing, and reports that fruit set requires pollen from a plant of a different compatibility type. A robusta seedling is therefore not a copy of its mother. To keep a good tree, it is propagated from cuttings as a clone, and because a clone cannot pollinate itself a planting needs several clones that are compatible with each other. The review describes the Brazilian state institute's method for conilon: ten steps taking twelve years, ending in clonal cultivars that are released and planted in sets.

The same review records the alternative: synthetic varieties, grown from seed harvested from a garden of selected clones allowed to cross. These are cheaper to distribute and less uniform. It names three genetic groups from which current robusta varieties derive: Congolese, Guinean, and the conilon group grown in Brazil, which is presumed to have arisen from a cross between the first two.

From a cross to a variety: three routes

CoffeeHQ explanation

A diagram of the logic, drawn by CoffeeHQ from the two breeding reviews. It is a model of the routes and gives no timings, because the reviews give them for particular programmes only.

F1 hybrids, and what hybrid vigour amounts to

Sourced

An F1 hybrid is the first generation from a cross between two distinct parents. The 2026 review traces how arabica breeders came to them late. A study of 1950 found no yield advantage in hybrids, but it had worked inside the narrow Typica and Bourbon stock, where the parents hardly differed. Vigour appeared once parents from outside that stock were used: in crosses of Ethiopian trees with Caturra derivatives, in programmes in East Africa and Cameroon, where what the review reports is better resistance to leaf rust and berry disease, and from 1990 in a Central American programme that crossed introgressed lines with Ethiopian trees.

On the size of the effect the review is careful, and the care matters. At research stations, a hybrid compared with the better of its own parents can sometimes yield more than three times as much; the authors say that such figures partly reflect how weak many inbred parent lines are. Under farmers' conditions and against the best commercial pure lines, they put the average gain across environments at 30 to 40 per cent. Those are a review's summary figures from trials largely run by its authors' own institutions, which have bred and promoted these hybrids. They are evidence from experiments, reported by interested parties.

Independent of that, a variety trial coordinated across many countries and published in 2025 scored 29 varieties for leaf rust at 23 sites. The variety with the least rust across sites was an F1 hybrid, and the most affected was Pacamara. The trial was not designed to compare genetic backgrounds, but its authors report that varieties carrying introgressions from other species had less rust on average than pure arabicas. That paper reports rust scores and not yield.

Why a species with so little diversity shows hybrid vigour at all is not settled. The review proposes two mechanisms, complementary blocks of introgressed robusta material and a rebalancing of how the two halves of the genome are regulated, and labels the second explicitly as a proposal. It also names the practical limit: hybrid plants are multiplied in the laboratory from tissue, which needs quality control to keep them true, or by seed from male-sterile mother plants, of which very few validated sources exist.

Resistance breeding, and how long resistance lasts

Sourced

Resistance to leaf rust in coffee is described in the breeding literature as governed by at least nine major genes, numbered SH1 to SH9. A 2023 paper on resistance breeding gives their origins: four identified in arabica itself, one derived from Coffea liberica, and four identified in robusta and present in descendants of the Timor Hybrid. It adds that the fungus is known to carry a matching virulence gene for each. The rust fungus exists as many races, more than fifty on the count of a review described in the guide to pests and diseases, and a variety is resistant only to races that cannot overcome the genes it carries. When a race that can do so becomes common, the variety's resistance is said to have broken down. The gene has not changed. The fungus population has.

This is why resistance is not a permanent property of a variety name, and why the 2025 trial is worth reading closely: every variety showed some rust symptoms at some site, and the ranking of varieties changed from one environment to another. Its authors call for continuous surveillance to detect breakdown in commercial varieties.

The breeder's answer is to combine several resistance genes in one plant, so that a race must overcome all of them at once. This is called pyramiding. A study from Brazil, published in 2023, shows what it involves: twelve crosses, 144 plants, and DNA markers linked to a rust-resistance gene and to a gene for resistance to coffee berry disease, used to pick out the eleven plants that carried the combination. That is a demonstration that the method works in a breeding population. Whether resistance built that way lasts longer in farmers' fields is a claim about the future, and the papers read here state it as an expectation.

Marker-assisted and genomic selection: what has been shown

Sourced

Marker-assisted selection uses a DNA test in place of waiting for a trait to show. It works where a trait is controlled by one or a few known genes, as in the pyramiding study above: a seedling can be tested for a resistance gene years before it would meet the disease in a field.

Genomic selection is a different idea. For traits controlled by many genes of small effect, yield above all, thousands of markers across the genome are used together to predict a plant's breeding value from a statistical model trained on plants that have been both genotyped and measured. The first such study in arabica, published in 2018, used 195 plants from thirteen families, about 21,000 markers and three years of field data on eighteen traits. Its authors conclude that selection cycles could be halved.

That is one population in one programme, and the result is a prediction accuracy inside that population. It does not show that a variety has been bred and released faster by this means. The 2026 hybrid review describes what genomic prediction can do for choosing the parents of a hybrid, and in the same breath calls its own framework a conceptual model that has yet to become a predictive breeding strategy. A reader who sees genomic selection described as how coffee is now bred should ask which variety it produced.

Variety, cultivar, line, population, group, clone

CoffeeHQ explanation

These words are used loosely on bags and fairly consistently in the breeding papers read for this page. The definitions below are CoffeeHQ's summary of how those papers use them, not a formal standard, which was not consulted.

Variety and cultivar

Used interchangeably in the coffee papers read here for a named, distinguishable kind of plant in cultivation. Botany reserves variety for a rank below species, as in Coffea liberica var. dewevrei; in coffee breeding and the trade, variety almost always means cultivar.

Pure line

A variety of a self-fertile species that has been self-pollinated until its offspring are uniform and breed true. Most traditional arabica varieties are pure lines, and their seed can be replanted.

F1 hybrid

The first generation of a cross between two distinct parents, distributed as that generation. Uniform in the field, but its seed segregates into unlike plants, so it has to be bought again or copied.

Clone

Plants multiplied vegetatively from one individual and genetically identical to it. The normal form of an improved robusta, planted in sets of mutually compatible clones.

Synthetic or composite variety

Seed harvested from a group of selected parents left to cross among themselves. Used in robusta where clones are too costly to distribute; less uniform than a clone by design.

Population or landrace

A mixture of related but genetically distinct plants maintained by farmers or surviving in forest, with no single pedigree. Much Ethiopian coffee is of this kind, which is why a variety name is often not available for it.

Group or lineage

A set of varieties sharing an origin: the Typica and Bourbon lineages in arabica, the Congolese, Guinean and conilon groups in robusta, or the families descended from the Timor Hybrid. A group is a statement about ancestry and says little about any one variety's behaviour.

Accession

One catalogued sample in a collection. Genome studies report results for accessions; whether an accession represents a variety as farmers grow it is a separate question.

The other species, and a caution about them

Sourced

Two species supply almost all the world's coffee. A third, long treated as Coffea liberica with two varieties, Liberica and excelsa, is a small share of production. A study published in Nature Plants in 2025, using genomic data together with the form of the plants and where they grow, supports dividing it into three species: Coffea liberica, Coffea dewevrei for excelsa, and a little-known third, Coffea klainei. With those reinstated its authors count 133 species of coffee. The same paper says excelsa is drawing attention in Uganda, South Sudan and Guinea because it can produce a commercial crop at higher temperatures and through longer dry periods than robusta. That is its authors' description of why the crop is being planted, not the result of a comparative trial reported in that paper.

Coffea stenophylla, a West African species with black fruit that was farmed on a small scale more than a century ago, was found again in the wild in Sierra Leone in 2018; it had not been recorded there since 1954. The study that reports this describes it as extremely localised and seemingly threatened in both places it was found. A later study of its chemistry describes it as an undomesticated species of commercial interest for its heat tolerance and arabica-like flavour, and gives the comparison behind the first: it grows naturally where the mean annual temperature is 6.2 to 6.8 degrees Celsius higher than where arabica does.

None of this makes any species proof against a changing climate. Growing naturally in a warmer place is an observation about where a wild plant survives. It is not a measurement of yield, of quality under farm conditions, or of how the plant responds to drought, disease or a season hotter than it evolved in. Stenophylla has no released variety, and no yield figure for it was read. The claim that can be supported is narrower: breeders have more to work with than two species, and the wild populations holding that material are small.

What is not said here about genes, flavour and editing

CoffeeHQ explanation

No variety is recommended, and no claim is made that one variety tastes better than another: flavour depends on where and how a tree is grown and processed as well as on its genes, and the variety pages give what trials found. Genetic modification and gene editing are not covered beyond the hybrid review's remark that editing has been shown to work in coffee and that a commercially viable way of making male-sterile plants with it remains a prospect. The genetics of caffeine, of cup quality and of drought tolerance are not described, because the studies read here do not establish them. Dates for the origin of arabica are given as the modelled ranges they are.

What to do next

• The Timor Hybrid — The plant itself: where it was found, what was bred from it, and what trials of its descendants report.
• F1 hybrids — Named hybrids, what the catalogue records about each, and why they cannot be grown from saved seed.
• Robusta cultivars — How a self-incompatible species is released as sets of clones, with the programmes that did it.
• Coffee pests and diseases — The rust fungus, its races, and the other organisms resistance breeding is aimed at.
• Ethiopia in the atlas — Where wild arabica grows, how exposed it is, and why so much of it has no variety name.