Genetic diversity

August 22, 2026August 23, 2026

50 years of cassava diversity: what went into the bank, and what came out

Two new papers in Plants, published two months apart, tell the story of CIAT’s cassava genebank ¹ from opposite ends: how the collection was assembled and conserved, and what breeders have actually done with it. Read together, they amount to a remarkably candid 50-year audit of a slow-motion agricultural asset. I’ll just give you the main beats here. It’s really worth reading both papers in full.

Part I is the origin story. Botanist Victor Manuel Patiño’s 1969–70 expeditions alone brought in roughly a third of today’s 5,000 or so cassava landraces, transported as stem cuttings across Colombia, Ecuador, Venezuela and beyond, often in difficult conditions and under changing quarantine rules.

The collection has its blind spots: Brazil and the Guianas remain poorly represented ², passport information can be patchy, and the overwhelmingly male composition of historical collecting teams probably meant that some of the varietal knowledge held by women farmers went unrecorded.

Conservation itself has been a moving target. It has evolved from field genebanks to in vitro slow-growth tissue culture storage after a frogskin-disease scare forced the field collection to close in 2003, and now cryopreservation at CIAT’s Future Seeds facility. Meanwhile, DNA fingerprinting keeps revealing an awkward truth familiar to genebank curators everywhere: the names people give varieties and the genetic identities of the material do not always agree. A lot of effort over the years has also gone into testing for pathogens to ensure that distribution is safe.

Part II asks what all this diversity is actually for? The answer has also changed considerably over five decades. Early researchers chased traits such as high protein and low cyanogenic content before turning towards yield and starch percentage in the Green Revolution era.

The payoffs have been considerable: landraces have contributed traits such as resistance to pests and diseases, adaptation to acid soils and highland cold, and quality traits for fresh-market cooking versus industrial starch. The route from accession to released variety is rarely direct: a landrace gets screened for a trait, crossed, then recombined and selected over several more generations before anything reaches a farmer’s field. The results include varieties like Nataima-3, bred for whitefly resistance thanks to an Ecuadorian landrace. At the same time, the authors argue that cassava breeding now needs to move beyond broad phenotypic selection towards more systematic use of inbred lines, because the crop’s high heterozygosity makes the introduction of specific traits particularly difficult.

The two papers therefore tell a big story about genebanks. Collecting diversity is only the beginning; its value is realized decades later, when a breeder encounters a problem that nobody could have anticipated when the stuff was first collected. The cassava collection is a long-term portfolio of biological options, one whose contents have taken half a century to assemble, whose inventory is still being corrected, and whose most valuable assets may be the ones that have not yet been used.

A new project is looking to genetically engineer cassava to photosynthesize more efficiently at higher temperatures. I do wonder whether someone has already checked whether any of those 5,000 landraces might help with that. Maybe they can’t, but it’s worth having a look. And you never know, something else of interest might jump out. That’s the beauty of large international collections of crop diversity such as CIAT’s.

August 21, 2026August 17, 2026

Conserving the tangle of grapevines

I think we may have already pointed to Conservation gap analysis for wild grapevines (Vitis L.) of the Americas, the latest in a series of papers by our friend Colin Khoury and a rotating assortment of colleagues on the conservation status of the crop wild relatives of the Americas, genepool by genepool. The authors compiled occurrence records for 38 wild American grapevine taxa, and used fancy GIS to infer the overall distribution and environmental niche of each. They then assessed the degree of representation of each taxon in genebanks and protected areas, and hence any remaining conservation gaps. Here’s the headline finding:

We categorize 25 of 38 of the taxa as urgent priority and 10 as high priority for improving ex situ conservation representation. Three taxa are assessed as urgent and 29 as high priority for enhancing in situ conservation. Further action, with emphasis on conservation gap hotspots, is needed to more comprehensively conserve wild Vitis native to the Americas.

Which is pretty clear.

Or is it?

What if “taxa” are perhaps not always the best units of conservation to use in assessing conservation efforts?

That may in fact be one of the implications of a paper that came out just a few weeks after that of Colin and friends: The dynamics of introgression and parallel adaptation across North American Vitis species.

These authors show that introgression and hybridization are pervasive and evolutionarily important across North American Vitis, based on genomic analysis of 639 accessions representing 48 species. About 14% of the average genome shows evidence of introgression, particularly associated with areas where species come into contact. Some taxa usually regarded as hybrid species are in fact better understood as ever-changing hybrid swarms, rather than distinct evolutionary lineages. Most importantly, the authors find that introgressed genetic variants have repeatedly contributed to adaptation in different species. The paper therefore portrays Vitis diversity as a reticulate network — or tangle — of species, populations and gene flow, rather than a set of discrete species.

This has important implications for conservation: hybrid zones and admixed populations may be really significant reservoirs of adaptive diversity. The framework of the first paper might potentially underestimate the conservation importance of regions where these occur, if they contain substantial genetic variation but aren’t well represented by the taxonomic units used in the gap analysis. For example, it might happen that two neighbouring species are reasonably well represented ex situ, but not from the specific regions where they hybridize and introgression occurs.

This suggests a useful next, synthetic step: take the geographic gaps from the first paper and overlay them with the evidence for introgression and gene flow networks from the second. The resulting map could identify not just under-collected species, but under-collected (or under-conserved in situ) evolutionary processes and genetic mixtures. That could be valuable for designing the next Vitis collecting mission.

August 19, 2026August 18, 2026

The revenge of the sweetpotato

Speaking of digital imagery and its uses in genebanks, get a load of the recently published catalogues of the Cuban sweet potato collection at the Research Institute of Tropical Roots and Tuber Crops (INIVIT), and of Peruvian cacaos. Beautiful. And do yourself a favour and don’t skip the foreword, preface and introduction to the Cuban volume. I found them very eloquent — and moving. Maybe a touch overwrought, admittedly, but it does take some gumption to say, of sweetpotato, in a genebank catalogue of all places, that: “Today, history grants it its revenge.” I just hope it’s true.

August 18, 2026August 19, 2026

A picture is worth a thousand descriptors

For decades, germplasm characterization has relied on people looking at plants, seeds and fruits and recording what they see; first on paper forms, more recently admittedly on tablets and the like. I’ve done that myself, and let me tell you, recording the colour of taro stems on bits of damp paper in the middle of a forest clearing in Vanuatu is no fun.

Lately, thankfully, the camera has been taking over.

A recent overview of new tools for plant genebanks highlights digital photography as a way of capturing standardized information on the colour, size and morphology of seeds and other plant parts, alongside more sophisticated technologies such as hyperspectral imaging and mobile field sensors. The attraction is obvious: instead of recording a handful of descriptors by eye, images can capture a much richer set of characteristics that can subsequently be measured and analysed at your leisure.

The potential is particularly striking for fruit crops. A new study of heritage apples used controlled multi-view imaging to characterize about 350 accessions over three years and two locations. Fancy maths achieved 95% accuracy in distinguishing 38 cultivars, rising to 99% when images of three fruits were combined.

And the technology does not necessarily require sophisticated equipment. In a recent demonstration with beans with complex colour patterns, Miguel Angel Acosta Chinchilla used ordinary photographs, image pre-processing and clustering algorithms to extract dominant palettes and colour distributions.

We’re moving from a limited number of human-defined descriptors to infinitely explorable machine-readable phenotypes. An image can preserve information that nobody thought to score at the time, and algorithms can return to it later to measure traits that were not originally part of the characterization protocol. For genebanks, that could be transformative. A photograph taken today may become a source of data for questions that don’t actually arise until tomorrow.

I wish I had a digital camera with me in that taro patch twenty-odd years ago. Goodness knows what people could be finding in those photos now.

August 17, 2026August 17, 2026

Brainfood: Mung bean pan-genome, Sunflower resistance, Olive adaptation, Allergic peanuts, Weird coffees, Chinese tea, Measuring selection

Graph-based pan-genome reveals structural variations associated with agronomic traits in mung bean. Pan-genome captures extensive structural variation missed by a single reference genome, providing new targets for mung bean improvement.
Wild genes to the rescue: high-throughput genomics reveals the wild source of broomrape resistance in sunflower. Combining high-throughput root phenotyping with genomics identifies wild introgressions underlying broomrape resistance in sunflower and exposes valuable adaptive variation for breeding.
Genomic insights into the local adaptation of spontaneous olive trees in the Mediterranean Basin highlight the need for conservation planning in the face of global change. Genomics of wild West Mediterranean olive trees reveals patterns of local adaptation across temperature and rainfall gradients that it will be important to conserve as climates change.
Genetic regulation of major immunogenic protein accumulation in peanut seeds. Genomics provides targets for breeding peanuts with reduced allergenicity.
Genomic elucidation of hybridization between Liberica and excelsa coffee and its implications for coffee crop development. Genomics confirms that Liberica and excelsa coffee readily hybridize and produce fertile, genetically diverse offspring with intermediate agronomic traits, a possible source of diversity to broaden coffee’s climate resilience.
Simple sequence repeat (SSR) marker-based genetic diversity and population structure of large-leaf tea germplasm across four major tea-producing regions of Yunnan, China. Genomics of Yunnan’s large-leaf tea germplasm finds substantial genetic diversity but remarkably little regional differentiation, showing that seed propagation, natural crossing and human movement have maintained genetic connectivity across tea-growing regions.
Measuring maternal line selection driven by differential survival in ex situ collections for plant conservation. A new method distinguishes random mortality from selection among maternal lines in ex situ collections, offering conservationists a way to detect whether plant deaths are systematically eroding the genetic variation they are trying to preserve. And no genomics involved.