Selective sweep
More actions
A selective sweep is the loss of genetic diversity around an allele that selection carries quickly to high frequency. Neutral variants on the stretch of chromosome where the favoured allele arose rise in frequency with it while the other variants at those sites fall. Farther from the selected site, recombination separates more of the neighbouring variants from the favoured allele before it spreads, so less diversity is lost.[1] The width of the affected region depends on the strength of selection relative to the rate of recombination.[2][3] Genome scans search for such regions to find the targets of selection. Changes in population size, such as a population bottleneck, can leave similar patterns.[4] So can selection against harmful mutations.[2]
In Cannabis sativa, scans of resequenced genomes have reported candidate sweep regions between hemp and drug-type plants and between cultivated and wild plants. Guangpeng Ren and colleagues kept only the windows of the genome that stood out on three statistics together. Against a basal group made up mostly of feral plants and landraces from China, they found 671 such segments in hemp-type plants and 936 in drug-type plants. The genes in these regions are candidates, annotated from their closest matches in Arabidopsis.[5]
The hitch-hiking effect
In 1974 John Maynard Smith and John Haigh worked out how the substitution of a favourable allele changes allele frequencies at closely linked loci. Alleles on the chromosome on which the new mutation first occurred increase in frequency and the others decrease. They called this the hitch-hiking effect, since an allele can get "a lift in frequency" from selection on its neighbour. Average heterozygosity at a linked neutral locus falls by an amount that depends on its distance from the selected locus. Summing over many substitutions, they estimated that this effect would outweigh random fixation in populations of a million or more, and perhaps of 10,000 or more. They offered it as a reason why the amount of polymorphism differs less between natural populations than their sizes would predict.[1]
Recurrent sweeps predict lower diversity in three places: where recombination is restricted, where sweeps have been many and around the targets themselves. Each of these patterns has been found. From the late 1980s polymorphism was found to be reduced in parts of the genome with little crossing-over. David Begun and Charles Aquadro then showed that nucleotide diversity in Drosophila melanogaster rises with the rate of crossing-over.[2]
Signs in the genome
Under the classical model a single new mutation rises under constant selection in a randomly mating population of constant size. Its footprint has four parts:[6]
- little polymorphism around the selected site;
- an excess of rare variants, at the site and in the regions on either side;
- an excess of high-frequency derived variants (the newer state at a site), in the flanking regions only;
- strong linkage disequilibrium within each flank, with none between variants on opposite sides of the selected site.
Tests built on these parts look at different features of the data: the level of diversity, the frequency spectrum of variants and the haplotypes around a site.[2] The signals fade at different rates. Drift removes ancestral haplotypes, while recombination and new mutation break up the swept one. A skew towards a few common haplotypes therefore lasts only about 0.01 Ne generations, where Ne is the effective population size. That is an order of magnitude shorter than the signal in the frequency spectrum.[2]
While a favoured allele is still rising, it climbs so fast that it sits on an unusually long haplotype of low diversity, while chromosomes without it look like the rest of the genome. Benjamin Voight and colleagues built a statistic on this contrast and applied it to about 800,000 SNPs in 309 people from three samples of the HapMap project. A region holding the lactase gene LCT was among the clusters of extreme values in Europeans.[7]
Where one group went through a sweep and another did not, allele frequencies at and around the selected site differ between them. The cross-population composite likelihood ratio test (XP-CLR) uses this differentiation at linked loci. The fixation index FST and the ratio of diversity between the groups measure it more simply.[5]
One haplotype or several
In a hard sweep a new advantageous mutation rises rapidly to fixation and carries with it part of the haplotype on which it arose. A soft sweep starts in one of two other ways. An allele already present on several haplotypes may become favoured, for example after a change in the environment. It then carries all of them up. Alternatively the same advantageous mutation may arise more than once. An allele that rises without reaching fixation still removes some linked diversity. Booker and colleagues call a sweep that is still polymorphic when sampled incomplete. They call one whose allele becomes effectively neutral partway a partial sweep.[2]
Pleuni Pennings and Joachim Hermisson modelled the same beneficial mutation arising repeatedly at realistic rates. The level of polymorphism and the frequency spectrum departed much less from neutral expectations than in the classical sweep, while linkage disequilibrium showed a stronger signal.[6] Statistics built on haplotypes detect soft sweeps better than those built on diversity or the frequency spectrum.[2]
Nandita Garud and colleagues scanned 145 sequenced strains of D. melanogaster from North Carolina with a haplotype statistic of their own. All of the 50 strongest candidate regions held several haplotypes at high frequency, a pattern their simulations generated much more easily by soft sweeps than by hard ones.[8] Rebecca Harris, Andrew Sackman and Jeffrey Jensen disputed this result and two others like it, on humans and on HIV. They found the methods prone to very high rates of false positives, reporting soft sweeps under models of hard sweeps and of neutrality. In their view well-fitted demographic histories with rare hard sweeps are the more parsimonious explanation.[4]
Other causes of low diversity
Selection against harmful mutations also removes the neutral variants linked to them. This background selection lowers local diversity and skews the frequency spectrum towards rare variants, as recurrent sweeps do. It can also account for the link between diversity and recombination in Drosophila. At scales above about 100 kb it appears to be a major cause of differences in diversity along the genome in humans and D. melanogaster.[2]
A bottleneck in the group that carries an allele also lowers its diversity, but across the whole genome. Comparing the region with many others in the same samples tells the two apart.[3] Harris and colleagues criticise a tendency to fit sweep models to genomic data without proper demographic null models.[4]
Domestication was expected to bring a small population under the stresses of selection, and with it reduced diversity, sweeps and a heavier load of harmful mutations. All three have been observed. Robin Allaby and colleagues set this against archaeological evidence from cereals. Selection on traits such as the loss of seed shattering was weak, comparable to natural selection. Domesticated forms emerged slowly. On their reading few loci were under selection, which left few occasions for hitch-hiking.[9]
Sweeps under domestication
Artificial selection in rice, Oryza sativa, left a sweep around the Waxy gene, which encodes an enzyme of starch synthesis. A change at the splice site of its first intron drastically reduces amylose in the grain of glutinous rice. Kenneth Olsen and colleagues found the change in 17 of the 20 typed accessions of temperate japonica, the low-amylose rice of north-eastern Asia, against 2 of 20 indica and 1 of 17 tropical japonica. In the gene, silent-site diversity among carriers was 97 per cent lower than among non-carriers. Across 111 gene fragments from the rest of the genome the ratio of carriers' diversity to non-carriers' averaged 0.46, against 0.03 at Waxy. Tajima's D, a summary of the frequency spectrum, reached its lowest value in the region at the gene. Carriers shared a long block of linked variants around the mutation, while in non-carriers linkage disequilibrium fell away quickly on either side.[3]
Diversity was measured at 12 genes spread over 500 kb around Waxy. The valley of low diversity in carriers spanned about 250 kb and at least 39 genes. From its width Olsen and colleagues estimated a selection coefficient of about 4. Their estimate is higher than those from sweeps in wild species and at maize domestication genes, which were also made from the size of the swept region. Rice mostly fertilises itself, which lowers effective recombination and widens a sweep for a given strength of selection, though the smaller effective population size of an inbreeder also weakens selection against drift.[3]
Scans in cannabis
Scans in cannabis compare groups of plants defined by use or by origin, divide the genome into windows and keep the windows with the most extreme values of one or more statistics.
| Study | Plants | Groups compared | Statistics and window | Windows kept | Reported |
|---|---|---|---|---|---|
| Ren and colleagues (2021) | 104 accessions from around the world | Hemp-type and drug-type plants, each against a basal group; hemp against drug-type cultivars | FST, ratio of diversity, XP-CLR; 10 kb | Top 5 per cent on all three | Against the basal group, 671 segments (510 genes) in hemp-type and 936 (689 genes) in drug-type plants[5] |
| Chen and colleagues (2022) | 21 Chinese accessions, with 26 published Chinese and 5 foreign genomes | Cultivated against wild | FST, ratio of diversity; 10 kb | Top 1 per cent on both, autosomes and X chromosome apart | 209 genes, nine of them related to flowering[10] |
| Lynch and colleagues (2025) | Chromosome-scale assemblies (78 haplotypes) | Marijuana-type (high-THC) against hemp | XP-CLR; 20 kb | Top 5 per cent | GIGANTEA in a significant region of the X chromosome[11] |
| Wang S. and colleagues (2025) | 113 accessions (109 kept), 110 of them genomes published earlier | Hemp-type cultivars against basal feral plants; drug-type cultivars against drug-type feral plants | XP-CLR, FST | Top 10 per cent on both | 1,182 and 537 genes[12] |
| Wang B. and colleagues (2026) | 271 accessions from 29 sites in 12 provinces of China | Three pairs of regions of China | XP-CLR; 100 kb | Top 5 per cent | 3,299, 3,215 and 3,404 genes[13] |
Hemp and drug types
Ren and colleagues resequenced cannabis from across its range and found four genetic groups. A basal group of 14 feral plants and landraces from China and two feral plants from the United States is sister to the rest, a hemp-type group and two drug-type groups.[5] In every group linkage disequilibrium fell to half its value within 10 kb, which set the size of the windows. The authors note that diversity and FST are the measures most used in crops and domestic animals. To tell sweeps from background divergence caused by bottlenecks they also required XP-CLR. Only windows in the top 5 per cent on all three counted as putative sweeps. They also note that the groups are not randomly mating populations: apart perhaps from the feral plants, each was bred separately at a probably small effective size.[5]
Hemp-type and drug-type plants were each compared with the basal group, which may stand for an early stage of domestication. Of the genes in these windows, 253 were shared by the two comparisons. Among the genes the authors highlight in hemp-type cultivars were genes that inhibit branching (D14, KNAT1), genes of flowering time and photoperiod response (FLK, EHD3) and genes of cellulose and lignin synthesis (SS, SPS1). In drug-type cultivars they highlight genes that promote branching (NDL2, DTX48), genes of flowering time (HUA2, FPF1) and genes of lignin synthesis (CSE, C4H). Against hemp-type cultivars, drug-type cultivars also showed selection on HDR, a gene of the pathway that supplies substrates for cannabinoid synthesis. No functional category of genes was significantly enriched after correction for multiple tests.[5] In a separate analysis, which mapped reads to a reference genome carrying both synthase genes, complete THCA synthase sequences were largely missing from hemp and complete CBDA synthase sequences from drug-type cultivars.[5]
Shuyu Wang and colleagues added three cultivars from north-eastern China to 110 genomes from around the world published earlier and built a pangenome, keeping 109 accessions for the scan. Genes in the windows between hemp-type cultivars and basal feral plants were enriched for flower development and for regulators of transcription. Among them were genes of flowering (Hd3a, HDR1, FPF1) and genes the authors relate to fibre (KCS, GA20-oxidase1, TBL). The windows between drug-type cultivars and drug-type feral plants showed no significant enrichment. They held flowering genes such as SPL and LATE. The authors relate SPL to the growth of glandular trichomes from work in other plants.[12]
Ryan Lynch and colleagues built a pangenome from 193 genomes. Among its chromosome-scale assemblies, the mean FST between marijuana-type (high-THC) and hemp populations was 0.20, a level of differentiation they compare with that between species of strawberry. A SNP in GIGANTEA, a gene of the circadian clock that acts on flowering time and cell elongation, had the fifth-highest FST, 0.77. An XP-CLR scan in 20 kb windows also placed the gene in a significant region of the X chromosome. The authors relate it to the differences in flowering and internode length between hemp and marijuana-type plants.[11]
Wild and cultivated plants in China
Xuan Chen and colleagues compared cultivated with wild cannabis from China using FST and the ratio of diversity in wild to cultivated plants, after removing outliers and admixed plants. They scanned the X chromosome apart from the autosomes, because they regard it as more sensitive to domestication history and selection. The 209 genes in the top windows of both statistics they call positively selected genes. Nine relate to flowering, among them FY, PRR37 and a homologue of CENTRORADIALIS. Seven relate to seed germination and plant development.[10]
Baishi Wang and colleagues resequenced 271 plants from 29 sites in 12 provinces of China, with 6 to 30 plants from each site. They ran XP-CLR on three pairs of populations: northern against southern Xinjiang; Xinjiang against Tibet (Xizang); and south-western China (Tibet, Sichuan, Yunnan) against north-eastern China (Inner Mongolia, Heilongjiang). The three comparisons gave 3,299, 3,215 and 3,404 genes in significant windows, with few genes in common.[13] For 90 plants from 14 populations they also measured cannabinoids, split the plants into high- and low-content groups and compared the groups with FST alone. They call the resulting regions, 41.2 Mb with 1,724 genes for CBD and 40.2 Mb with 1,238 genes for THC, chemotype-associated selective sweeps. Six of the genes were annotated as CBDA synthase or THCA synthase.[13]
See also
References
- ↑ 1.0 1.1 Maynard Smith, J.; Haigh, J. (1974). "The hitch-hiking effect of a favourable gene". Genetical Research. 23 (1): 23–35. doi:10.1017/S0016672300014634.
- ↑ 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 Booker, T.R.; Jackson, B.C.; Keightley, P.D. (2017). "Detecting positive selection in the genome". BMC Biology. 15 (1) 98. doi:10.1186/s12915-017-0434-y. PMC 5662103.
- ↑ 3.0 3.1 3.2 3.3 Olsen, K.M.; Caicedo, A.L.; Polato, N.; McClung, A.; McCouch, S.; Purugganan, M.D. (2006). "Selection under domestication: evidence for a sweep in the rice Waxy genomic region". Genetics. 173 (2): 975–983. doi:10.1534/genetics.106.056473. PMC 1526538.
- ↑ 4.0 4.1 4.2 Harris, R.B.; Sackman, A.; Jensen, J.D. (2018). "On the unfounded enthusiasm for soft selective sweeps II: examining recent evidence from humans, flies, and viruses". PLOS Genetics. 14 (12) e1007859. doi:10.1371/journal.pgen.1007859.
- ↑ 5.0 5.1 5.2 5.3 5.4 5.5 5.6 Ren, G.; Zhang, X.; Li, Y.; Ridout, K.; Serrano-Serrano, M.L.; Yang, Y.; Liu, A.; Ravikanth, G.; Nawaz, M.A.; Mumtaz, A.S.; Salamin, N.; Fumagalli, L. (2021). "Large-scale whole-genome resequencing unravels the domestication history of Cannabis sativa". Science Advances. 7 (29) eabg2286. doi:10.1126/sciadv.abg2286.
- ↑ 6.0 6.1 Pennings, P.S.; Hermisson, J. (2006). "Soft sweeps III: the signature of positive selection from recurrent mutation". PLoS Genetics. 2 (12) e186. doi:10.1371/journal.pgen.0020186.
- ↑ Voight, B.F.; Kudaravalli, S.; Wen, X.; Pritchard, J.K. (2006). "A map of recent positive selection in the human genome". PLoS Biology. 4 (3) e72. doi:10.1371/journal.pbio.0040072. PMC 1382018.
- ↑ Garud, N.R.; Messer, P.W.; Buzbas, E.O.; Petrov, D.A. (2015). "Recent selective sweeps in North American Drosophila melanogaster show signatures of soft sweeps". PLOS Genetics. 11 (2) e1005004. doi:10.1371/journal.pgen.1005004. PMC 4338236.
- ↑ Allaby, R.G.; Ware, R.L.; Kistler, L. (2019). "A re-evaluation of the domestication bottleneck from archaeogenomic evidence". Evolutionary Applications. 12 (1): 29–37. doi:10.1111/eva.12680.
- ↑ 10.0 10.1 Chen, Xuan; Guo, Hong-Yan; Zhang, Qing-Ying; Wang, Lu; Guo, Rong; Zhan, Yi-Xun; Lv, Pin; Xu, Yan-Ping; Guo, Meng-Bi; Zhang, Yuan; Zhang, Kun; Liu, Yan-Hu; Yang, Ming (2022). "Whole-genome resequencing of wild and cultivated cannabis reveals the genetic structure and adaptive selection of important traits". BMC Plant Biology. 22 (1) 371. doi:10.1186/s12870-022-03744-0.
- ↑ 11.0 11.1 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.
- ↑ 12.0 12.1 Wang, S.; Zhong, X.; Cheng, Y.; Yu, Y.; Wan, J.; et al. (2025). "Pan-genome analysis of Cannabis sativa: insights on genomic diversity, evolution, and environment adaption". International Journal of Molecular Sciences. 26 (17) 8354. doi:10.3390/ijms26178354.
- ↑ 13.0 13.1 13.2 Wang, Baishi; Gao, Shan; Wang, Xue; Yun, Quanzheng; Nie, Hao; Li, Ning; Li, Yonggang; Feng, Deri; Li, Linsen; Li, Tianye; Lu, Xu; Liu, Jingjing; Wei, Xuge; Mu, Yanhong; Wu, Qiqi; Hu, Jinyuan; Liu, Fangru; Feng, Rui; Yin, Jianyu; Zhang, Qikun; Zhang, Ying; Ren, Fu (7 September 2026). "Genomic variation in 271 diverse accessions of Cannabis sativa L. in China". BMC Plant Biology. doi:10.1186/s12870-026-09909-5.