Talk:Epistasis

Talk

Discussion for: Epistasis • View • Links • Related changes

Use this page for verification notes, disputes, and action items. Keep narrative content on the article.

Page scope

Epistasis covers interaction between genes at different loci: masking in a cross, departure from additivity, epistatic variance, how interactions are found, what they mean for selection and hybrids, and what has been measured in cannabis. The relation between two alleles at one locus belongs to Dominance and recessivity and heterozygote advantage to Overdominance. One gene affecting several traits belongs to Pleiotropy, the performance of hybrids to Heterosis and the trait-by-trait inheritance of cannabis to Genetic traits.

Data verification

Record source conflicts, figures that differ between studies and sources not used here. Keep narrative content on the article.

Discussion

Sources

Phillips (2008), Nature Reviews Genetics: Bateson's and Fisher's senses, the three uses, ratios, sign epistasis, scale and the chicken selection lines. Crow (2010), Philosophical Transactions of the Royal Society B: the case that breeders can leave epistasis out. Mackay (2014), Nature Reviews Genetics: the case for pervasive epistatic gene action, detection and the real or apparent additive variance.
Hill and colleagues (2008), PLOS Genetics, and Mäki-Tanila and Hill (2014), Genetics: why variance is mostly additive. Bloom and colleagues (2015), Nature Communications, and Huang and colleagues (2012), PNAS: large panels in yeast and Drosophila.
Felsenstein (2019), Theoretical Evolutionary Genetics, pp. 363–364: epistasis for fitness and the curved scale. Beavis and colleagues (2023): the partition of genotypic variance.
Carlborg and colleagues (2006), Nature Genetics; Doebley and colleagues (1995), Genetics; Soyk and colleagues (2017), Cell; Caicedo and colleagues (2004), PNAS; Jiang and Reif (2015), Genetics; Bomblies and colleagues (2007), PLOS Biology: selection, background, prediction and hybrid necrosis in animals and plants other than cannabis.
Woods and colleagues (2021), Genetics; Stack and colleagues (2023), Horticulture Research; Welling and colleagues (2019), Scientific Reports; Campbell and colleagues (2020), Cannabis and Cannabinoid Research; de Meijer and colleagues (2009) and de Meijer and Hammond (2016), Euphytica; Carlson and colleagues (2021), Journal of Experimental Botany; Toth and colleagues (2022), Frontiers in Plant Science: cannabis.

Where the sources disagree

How much epistasis matters. Hill and colleagues and Crow treat additive variance as the working quantity for prediction and selection. Mackay (2014) calls epistatic gene action common and additivity possibly an emergent property of it. The two sides agree that most observed variance is additive. They differ on what follows from that.
The chicken network. Phillips (2008) describes Growth9 and five further regions and six genes in all. The abstract of Carlborg and colleagues (2006) describes a network of four interacting loci.
The gene at LG9.40. Woods and colleagues propose olivetol synthase as the candidate, on sequence differences and a functional assay. Stack and colleagues equate the region with de Meijer's locus O on the phenotypes and designed their O marker from the low-cannabinoid USO-31 olivetol synthase sequence. De Meijer and colleagues (2009) say only that O governs a reaction on the way to the phenolic precursors.
What had been shown before 2021. Woods and colleagues write that the literature on cannabinoid inheritance had shown only dominance and additivity. Welling and colleagues had published an epistatic model for side-chain length in 2019.
The inheritance of propyl share. De Meijer and Hammond (2016) propose a multiple-locus additive model from inbred lineages of six progenitors, with F₂ patterns compatible with oligogenic and polygenic segregation. Welling and colleagues (2019) accept digenic ratios within locus B genotypes in one F₂. They say lineage differences may account for part of the difference and that a polygenic model may still be correct across lineages.

How far the evidence goes

Stack and colleagues use the word "epistatic" for B and O and for the two Woods regions. The abstract of de Meijer and colleagues (2009) does not use it for O. The only variance share for an interaction of these loci is the Woods figure for the two regions, and the match between regions and loci is Stack and colleagues'. Woods and colleagues' share is from one F₂ of two hemp cultivars, on quantile-normalised trait values.
The 7:6:3 in the Welling study was tested after plants were sorted into classes by kernel density, in 50 plants, and its degrees of freedom are 2. A 9:3:3:1 was not accepted in the heterozygotes at locus B, and 7:6:3 could not be tested in them.
Campbell and colleagues compare models fitted to line means from four cohorts (two parents, F₁, F₂), and report moderate uncertainty in model selection. Their results do not locate any interacting loci. They add that more general epistatic effects were not considered.
The papers on flowering (Carlson and colleagues, Toth and colleagues) raise epistasis as a possibility. Neither fits an interaction term.
All the cannabis tests above use hemp cultivars or breeding lines. None of them uses a landrace population, and the statements on the page do not extend to one.
The general findings on the page come from yeast, Drosophila, mice, chickens, maize, tomato and Arabidopsis. Jiang and Reif's result for outcrossing species comes from maize and says nothing specific about cannabis.

Terms

"Hypostatic" is Bateson's term for the locus that is masked. "Epistacy" is Fisher's word, which did not last.
Phillips's "compositional" and "statistical" correspond approximately to other authors' "physiological" or "genetical" and "statistical". Phillips prefers not to use "physiological" or "functional" for substitution against a fixed background.
Positive and negative epistasis are Phillips's preferred pair: a phenotype higher or lower than expected. "Synergistic" and "antagonistic" mean different things in different papers, so the article avoids them.
LG6.35 and LG9.40 are Woods and colleagues' names for clusters of loci on their linkage map. Stack and colleagues place them on chromosomes 7 and 8 of the CBDRx reference genome.

Images reserved in the article

One image has a place in the article but is not yet uploaded:

File:Epistasis-nine-genotype-values.jpg (Departure from additivity), a figure to be drawn: three small line charts of the nine genotypes of two diploid loci, with parallel lines for no epistasis, spreading lines for a change in the size of effects and crossing lines for a change in direction, after the scheme of Mackay (2014). The values are illustrative and include no cannabis data.