Core collection
More actions
A core collection is a small subset of a gene bank's accessions, chosen to hold as much as possible of the genetic diversity of the whole collection with as little repetition as possible.[1] It usually takes in about 10 per cent of the accessions. The rest stay in store as a reserve.[2] Users can screen the core first and go on to the wider collection with what they learn from it.[3]: 88, 112
No core collection selected from the gene bank holdings of Cannabis had been published by November 2026. Matthew Welling and colleagues proposed one in 2016. In 2024 the curator of the hemp collection at Geneva, New York, said he wanted to create one.[4][5]
From one proposal to many subsets
Otto Frankel coined the term in 1984 for a set meant to "represent with a minimum of repetitiveness, the genetic diversity of a crop species and its relatives".[6] Collections had grown large enough to complicate both their upkeep and their use in breeding.[1] A. H. D. Brown set out a working method in 1989. His core held about 10 per cent of a collection, chosen from what was known of each accession's origin, genetic characteristics and possible value to breeders. Accessions were sorted into groups first, with each group contributing in logarithmic or absolute proportion to its size. Entries stayed separate, never bulked into one population, while the make-up of the core could change as new material or better data arrived.[2] By Brown's estimate, a 10 per cent sample of a collection of more than 3,000 accessions keeps about 70 per cent of the diversity.[7] The Food and Agriculture Organization's Global Plan of Action made core collections a priority in 1996, and procedures followed from the International Plant Genetic Resources Institute in 2000.[3]: 37
A tenth of a very large collection can still be too large for breeders to use. In 2001 H. D. Upadhyaya and R. Ortiz added a second stage at the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), where the chickpea holdings stood at 16,991 accessions. They evaluated a core of 1,956 for 22 morphological and agronomic traits and kept 211 of them as a "mini core".[8] ICRISAT has since sent mini cores of seven crops to research groups in 14 countries.[6]
Brown's core was a fixed set, revised over time. In 1995 M. C. Mackay argued for several sets built to different criteria, and the Centre for Genetic Resources in the Netherlands later let users select their own maximum-diversity subset online. Brown recommended naming a set made for one purpose after that purpose, as in "acid tolerant set".[1]
What a core is for
Core collections are not meant to replace the collections they come from.[3]: 112 A curator can use one to learn how diversity is spread through a collection, to find gaps and to divide responsibility among holders of the same crop.[3]: 36 For a user with no data on which accessions might be useful, screening a core usually costs less than screening everything, and the results point to a further subset.[3]: 88–89 A core can also set which accessions make up a gene bank's active collection, the stock kept ready for supply.[9] Demand for a core subset may rise substantially once all its accessions have been sequenced and the data are easy to reach, which has been shown for rice.[6]
The authors of a 2003 handbook on managing collections warned that cutting a whole collection down to raise its use threatens its future value, because breeding goals move. Grass breeding at Aberystwyth went from yield to quality, then to low-input systems, stress tolerance and amenity turf within a decade. The handbook's authors proposed pruning what is held ready for distribution and judging the conservation collection on conservation grounds alone, with a core on hand so that a user who asks for an accession not immediately available can be offered a genetically similar one.[3]: 88–90 At the International Potato Center in Lima, 15,000 clonal potato accessions came down to 3,500 genetically distinct cultivars once duplicates were eliminated, which reduced costs substantially. The handbook calls the result a type of core collection, though the duplicates were not kept.[3]: 53
How a core is chosen
Three kinds of data feed the choice: passport data on identity and origin, agronomic and morphological data, and molecular markers.[1] Selection often begins by dividing the accessions into similar groups and deciding how many to take from each, a step T. L. Odong and colleagues call arbitrary. A core meant for one kind of user can then be weighted towards a group, such as a crop's primary gene pool or the landraces of one country.[1]
Markers allow selection on the alleles themselves. In simulations of wild crop relatives published in 1993, D. J. Schoen and Brown found that collections of fixed size held more alleles when populations were assessed at marker loci first, particularly alleles confined to one area.[10] In the H strategy of Schoen and Brown, groups are sampled in proportion to the diversity within them. In their M strategy, the search is for the combination of accessions with the most alleles at the marker loci.[1] Software now runs such searches over large data sets: PowerCore applies an advanced form of the M strategy.[11] Core Hunter, released in 2009 and revised in 2018, can aim at allelic richness, which its authors call the taxonomist's preference, at representativeness, the breeder's preference, or at both.[12][9]
Cores also differ by purpose. Odong and colleagues divide them into three kinds, each with its own test.[1]
| Purpose | What a good core looks like | Measure proposed |
|---|---|---|
| Represent every accession | Each accession in the collection has a close match among the entries | Average distance from each accession to its nearest entry, as small as possible |
| Represent the extremes | Entries are as different from one another as possible | Average distance from each entry to its nearest neighbouring entry, as large as possible |
| Represent the distribution | Traits and regions occur in the core in the same proportions as in the collection | Quantile–quantile plots and the Kullback–Leibler distance |
Some cores are drawn from landraces alone. Among 388 glutinous rice landraces from the 'He' cultivation zone of Guizhou in China, 325 were Kam Sweet Rice, largely similar to one another, and 63 were other varieties with higher diversity. Wenhui Yang and colleagues compared five selection strategies on microsatellite data and settled on 65 entries, which held 90.86 per cent of the alleles.[7]
How a core is checked
A new core is compared with its parent collection or with cores made another way. Common measures include the difference in means and variances for each trait, the share of each trait's range retained, the Shannon index, the share of classes or alleles retained and chi-square tests on frequencies.[1] The chickpea mini core passed tests of this kind on means, variances and frequency distributions. The main correlations between its traits also survived the reduction.[8]
In their 2013 review, Odong and colleagues concluded that the choice of measure often follows earlier papers and has little to do with what the core is for. They argued that a core and its collection are a sample and its population, so that a two-sample comparison asks the wrong question. By their reasoning, similar Shannon values do not show quality: a good core should score higher than its collection, since it carries less redundancy.[1]
Odong and colleagues also asked for tests on data held back from the selection. A core built from half of a set of markers and tested on the other half shows how far it captures variation at loci it was not chosen on. In their test on common bean, cores chosen this way beat random selection by 5 to 15 per cent on one distance measure and 1 to 5 per cent on another, and coconut gave a similar result. A study of Medicago truncatula inbred lines found little gain over random sampling, which its authors put down to the lines having no redundancy. Cores built for the extremes did worse than random selection at representing every accession, and cores built to represent every accession did poorly on the measures for extremes. The review reads this as a sign of how hard it is to make one core serve every purpose.[1]
Cannabis
Welling and colleagues proposed in 2016 an international core of up to 400 cannabis accessions from diverse regions, drawn from existing gene banks. Theoretical studies of natural populations suggested that this number would hold 99 per cent of polymorphisms for alleles with a species-wide frequency above 2 per cent. Moving seed was the obstacle: publicly accessible seed banks were lacking, and parties to the Single Convention on Narcotic Drugs and the Convention on Psychotropic Substances faced limits on importing, exporting and transferring it. The authors therefore proposed a "virtual" core, with the accessions left where they were and described in one shared database. A first selection would then be tested empirically, with pairwise comparisons of allele frequencies deciding which accessions to add or drop. Public passport data for cannabis were then absent or of low quality.[4]
In 2022 Rachel Burton and colleagues, writing on hemp seed, still saw "a strong need for a core germplasm resource" for industrial hemp.[13] Zachary Stansell, curator of the Geneva collection, said in 2024 that he was very interested in creating a core that would capture hemp diversity with the fewest accessions, each as genetically distinct from the others as possible, with inbred lines derived from it used to map traits. He also planned to sequence every accession.[5] A United States Department of Agriculture handbook he co-edited, dated December 2025, states the aim to "designate and maintain a core collection of critical materials".[14]: 5 A public marker panel of 3,040 loci has been developed from the genomes of 92 accessions of the National Plant Germplasm System and from public sequence.[15] On 6 November 2026, 91 of the 696 accessions at Geneva could be requested.[16]
Welling and colleagues, Burton and colleagues, Stansell and the handbook all speak of a cannabis core as something still to come, and none names its entries. Other sets have been assembled for particular studies and are called panels by their authors: 123 hemp accessions for fibre traits, 84 accessions for an infrared test of cannabinoids and 72 accessions from the Ecofibre collection for mapping cannabinoid genes.[17][18][19] Marcello Borin and colleagues used the words "core collection" for 104 plants of 11 hemp varieties of high breeding value, supplied by one company to validate 20 markers. That set was chosen for origin and for being dioecious or monoecious, and did not come from a gene bank's holdings.[20] Selections kept inside company and breeders' collections would not appear in the published literature.
A core can represent only the collection it is drawn from, and sampling of the plant is itself uneven. R. C. Lynch and colleagues, analysing 193 genomes in 2025, found global cannabis diversity undersampled, with Asian germplasm notably underrepresented.[21] Coverage of the holdings by origin is treated under gene bank, and the storage and regrowing of accessions under ex situ conservation.
See also
References
- ↑ 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 Odong, T.L.; Jansen, J.; van Eeuwijk, F.A.; van Hintum, T.J.L. (2013). "Quality of core collections for effective utilisation of genetic resources review, discussion and interpretation". Theoretical and Applied Genetics. 126 (2): 289–305. doi:10.1007/s00122-012-1971-y.
- ↑ 2.0 2.1 Brown, A.H.D. (1989). "Core collections: a practical approach to genetic resources management". Genome. 31 (2): 818–824. doi:10.1139/g89-144.
- ↑ 3.0 3.1 3.2 3.3 3.4 3.5 3.6 Engels, J.M.M.; Visser, L., eds. (2003). A Guide to Effective Management of Germplasm Collections. IPGRI Handbooks for Genebanks No. 6. Rome: International Plant Genetic Resources Institute. ISBN 92-9043-582-8.
- ↑ 4.0 4.1 Welling, M.T.; Shapter, T.; Rose, T.J.; Liu, L.; Stanger, R.; King, G.J. (2016). "A belated green revolution for Cannabis: virtual genetic resources to fast-track cultivar development". Frontiers in Plant Science. 7 1113. doi:10.3389/fpls.2016.01113.
- ↑ 5.0 5.1 Schwartz, Jocelyn; Smart, Christine D.; Stansell, Zachary (January 2024). "The USDA-ARS Hemp Germplasm Repository". Cornell Hemp. Cornell University. Retrieved 5 November 2026.
- ↑ 6.0 6.1 6.2 Engels, J.M.M.; Ebert, A.W. (2021). "A critical review of the current global ex situ conservation system for plant agrobiodiversity. II. Strengths and weaknesses of the current system and recommendations for its improvement". Plants. 10 (9) 1904. doi:10.3390/plants10091904.
- ↑ 7.0 7.1 Yang, W.; Mao, M.; Qin, J.; Nasar, J.; Pan, Z.; Zhou, L.; Zhao, Q. (2026). "Genetic diversity and core collection construction of glutinous rice landraces in 'He' cultivation zone of Guizhou using SSR sequencing". PLOS One. 21 (3) e0343623. doi:10.1371/journal.pone.0343623.
- ↑ 8.0 8.1 Upadhyaya, H.D.; Ortiz, R. (2001). "A mini core subset for capturing diversity and promoting utilization of chickpea genetic resources in crop improvement". Theoretical and Applied Genetics. 102 (8): 1292–1298. doi:10.1007/s00122-001-0556-y.
- ↑ 9.0 9.1 De Beukelaer, H.; Davenport, G.F.; Fack, V. (2018). "Core Hunter 3: flexible core subset selection". BMC Bioinformatics. 19 203. doi:10.1186/s12859-018-2209-z.
- ↑ Schoen, D.J.; Brown, A.H.D. (1993). "Conservation of allelic richness in wild crop relatives is aided by assessment of genetic markers". Proceedings of the National Academy of Sciences. 90 (22): 10623–10627. doi:10.1073/pnas.90.22.10623.
- ↑ Kim, K.-W.; Chung, H.-K.; Cho, G.-T.; Ma, K.-H.; Chandrabalan, D.; Gwag, J.-G.; Kim, T.-S.; Cho, E.-G.; Park, Y.-J. (2007). "PowerCore: a program applying the advanced M strategy with a heuristic search for establishing core sets". Bioinformatics. 23 (16): 2155–2162. doi:10.1093/bioinformatics/btm313.
- ↑ Thachuk, C.; Crossa, J.; Franco, J.; Dreisigacker, S.; Warburton, M.; Davenport, G.F. (2009). "Core Hunter: an algorithm for sampling genetic resources based on multiple genetic measures". BMC Bioinformatics. 10 243. doi:10.1186/1471-2105-10-243.
- ↑ Burton, R.A.; Andres, M.; Cole, M.; Cowley, J.M.; Augustin, M.A. (2022). "Industrial hemp seed: from the field to value-added food ingredients". Journal of Cannabis Research. 4 (1) 45. doi:10.1186/s42238-022-00156-7.
- ↑ Stansell, Zachary; Barraco, Anthony; Osatuke, Anya; Evangelista, Roque; Brownstein, Korey; Hernandez, Kevin; Kraft, Sadie; Itokazu, Ana Gabriela; Gordon, Tyler, eds. (16 December 2025). USDA Hemp Phenotyping and Protocol Handbook, Version 4 (PDF) (Report). USDA Agricultural Research Service. Retrieved 7 November 2026.
- ↑ "Open Source Hemp Marker Panel". Breeding Insight. Retrieved 7 November 2026.
- ↑ "Accession search, taxon Cannabis, all statuses". GRIN-Global. USDA Agricultural Research Service, National Plant Germplasm System. Retrieved 6 November 2026.
- ↑ Petit, J.; Salentijn, E.M.J.; Paulo, M.-J.; Thouminot, C.; van Dinter, B.J.; Magagnini, G.; Gusovius, H.-J.; Tang, K.; Amaducci, S.; Wang, S.; Uhrlaub, B.; Müssig, J.; Trindade, L.M. (2020). "Genetic variability of morphological, flowering, and biomass quality traits in hemp (Cannabis sativa L.)". Frontiers in Plant Science. 11 102. doi:10.3389/fpls.2020.00102.
- ↑ Gloerfelt-Tarp, F.; Hewavitharana, A.K.; Mieog, J.; Palmer, W.M.; Fraser, F.; Ansari, O.; Kretzschmar, T. (2023). "Using a global diversity panel of Cannabis sativa L. to develop a near InfraRed-based chemometric application for cannabinoid quantification". Scientific Reports. 13 (1) 2253. doi:10.1038/s41598-023-29148-0.
- ↑ Welling, M.T.; Liu, L.; Kretzschmar, T.; Mauleon, R.; Ansari, O.; et al. (2020). "An extreme-phenotype genome-wide association study identifies candidate cannabinoid pathway genes in Cannabis". Scientific Reports. 10 (1) 18643. doi:10.1038/s41598-020-75271-7. PMC 7596533.
- ↑ Borin, M.; Palumbo, F.; Vannozzi, A.; Scariolo, F.; Sacilotto, G.B.; Gazzola, M.; Barcaccia, G. (2021). "Developing and testing molecular markers in Cannabis sativa (hemp) for their use in variety and dioecy assessments". Plants. 10 (10) 2174. doi:10.3390/plants10102174.
- ↑ Lynch, R.C.; Padgitt-Cobb, L.K.; Garfinkel, A.R.; Knaus, B.J.; Hartwick, N.T.; et al. (2025). "Domesticated cannabinoid synthases amid a wild mosaic cannabis pangenome". Nature. 643 (8073): 1001–1010. doi:10.1038/s41586-025-09065-0. PMC 12286863.