Talk:Genotype-by-environment interaction

Talk

Discussion for: Genotype-by-environment interaction • View • Links • Related changes

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

Page scope

Genotype-by-environment interaction covers the interaction itself: its forms, how it is measured and partitioned, what it means for selection, and what has been measured in cannabis. The ratio of genetic to total variance belongs to Heritability, the response of one genotype to its environment to Phenotypic plasticity, the inheritance of continuous characters to Quantitative trait, and how populations come to suit a place to Local adaptation.

Data verification

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

Discussion

Sources

• Beavis and colleagues (2023), "G x E", chapter 10 of Quantitative Genetics for Plant Breeding: the linear model, the Allard and Bradshaw types, the partition of the interaction variance, the reference population of environments, expected mean squares.
• Malosetti, Ribaut and van Eeuwijk (2013), Frontiers in Physiology 4: 44: the definition, norms of reaction, regression on the mean, AMMI, GGE biplots, factorial regression, QTL-by-environment models.
• Annicchiarico (2002), FAO Plant Production and Protection Paper 174: genotype and environment in yield trials, wide and specific adaptation, location and year interactions, static and dynamic stability.
• Felsenstein (2019), Theoretical Evolutionary Genetics, pp. 435–438: the interaction as a source of error in heritability; Feldman and Lewontin.
• Via and Lande (1985), Evolution 39: 505–522: the genetic correlation between states of a character.
• Campbell and colleagues (2019), Agrosystems, Geosciences & Environment 2: 1–11; Petit and colleagues (2020), Frontiers in Plant Science 11: 102; Ferfuia and colleagues (2021), Agronomy 11: 1424; Beleggia and colleagues (2023), International Journal of Molecular Sciences 24: 8969; Mahdi and colleagues (2026), Frontiers in Agronomy 8: 1745927; Nain and colleagues (2026), Industrial Crops and Products 250: 123890; Yazıcı and colleagues (2026), Crop Science 66: e70391; Tang and colleagues (2016), Industrial Crops and Products 87: 33–44; Solomon and colleagues (2026), Industrial Crops and Products 249: 123693: hemp trials.
• Naim-Feil and colleagues (2023), Plants 12: 493: clones of medicinal cannabis in one facility.
• Petit and colleagues (2020), Frontiers in Plant Science 11: 569958: association mapping across the three sites of the European panel.
• Halpin-McCormick and colleagues (2026), Journal of Heredity 117: 24–38: genomic estimates of adaptive value from public genome data.

Where the sources disagree

• Stem height in Petit and colleagues (2020). The abstract places most agronomic traits among those with large interactions. Table 7 gives stem height a genetic component of 6.84 per cent and an interaction of 2.31 per cent, a ratio of 0.34. The authors attribute such small ratios to the size of the environment component in those traits.
• The range of the ratio in the same paper. The text gives 1.42 to 5.41 for its four named traits. Table 7 also has 8.43 and 30.70 for the two fibre weights, 3.14 for decorticability and 1.61 for shives.
• First report. Petit and colleagues call theirs the first study of significant interactions in fibre and flowering traits in hemp. Campbell and colleagues, published the year before, report a significant interaction for days to maturity and stem diameter, in a different trait set and without fibre quality.
• Later sowing and seed yield. In Ferfuia and colleagues (2021) seed yield rose with later sowing and later flowering. Tang and colleagues and Campbell and colleagues report the opposite relationship. Ferfuia and colleagues ascribe the difference to the late dioecious cultivars in those trials, and to the cooler grain filling of a Belgian trial at a higher latitude.
• Mahdi and colleagues (2026). The abstract speaks of minimal variation in THC across environments and of the genetic stability of cannabinoid profiles. The discussion reads the variation in THC among cultivars across environments as a complex interaction. The paper also reads differences among cultivars within each site as indicating an interaction, and its tables give the tests site by site. Its 2023 data from Sikeston and Washington were lost to flooding, and the authors ask for more replication.
• Beleggia and colleagues (2023). Table 1 prints 17 degrees of freedom for genotype by year. Six genotypes and three years give 10.
• Nain and colleagues (2026). The abstract says five locations and names six.

How far the evidence goes

• Every multi-environment trial cited grew hemp. The medicinal cannabis study cited compares clones within one facility and gives no interaction term.
• The European panel of Petit and colleagues includes landraces and wild material among mainly fibre accessions, and its variance components cover the whole panel. No cited trial reports landrace populations as a group across environments. The landrace populations of Babaei and colleagues (2024) on Heritability were grown in one greenhouse.
• Each trial samples one kind of environment more than others. Campbell's cannabinoid and terpene figures compare two irrigation treatments at one site in one season, so they measure interaction with water and nothing wider. Beleggia's figures are genotype by year at one site. Petit's are genotype by location in one season, 2013, so site and year cannot be separated.
• The percentages rest on different calculations: sums of squares in Campbell, variance components in Petit. The two sets of numbers are not on one scale.
• Petit and colleagues' association study compares separate models for each site. Halpin-McCormick and colleagues predict adaptive values from genotype and the climate of georeferenced samples, 111 of the 149. Neither runs a QTL-by-environment test of the kind Malosetti and colleagues describe.

Terms

• G×E and GEI name the same thing. GE and GL or GY (genotype by location, by year) appear in Annicchiarico.
• Norm of reaction: the phenotype of one genotype as a function of the environment.
• AMMI: additive main effects and multiplicative interaction. GGE: genotype main effect plus genotype-by-environment interaction. WAAS: weighted average of absolute scores.
• Environment in the tables of the article means location in Petit, irrigation treatment and site in Campbell, a combination of site, year and sowing date in Ferfuia, and year in Beleggia.

Further figures from the sources

• Campbell and colleagues (2019), genotype, environment and interaction in per cent: stem diameter 36.1, 8.3, 14.9; stand establishment 24.2, 51.1, 6.6; carbon isotope ratio 23.9, 46.4, 28.9; cannabichromene 50.0, 0.0, 16.8; α-pinene 54.2, 2.7, 12.2; β-pinene 50.3, 0.2, 12.0.
• Ferfuia and colleagues (2021): 1000-seed weight 38, 24, 25; specific seed weight 3, 78, 9.
• Petit and colleagues (2020), genotype, location, interaction and the ratio: glucose 67.57, 11.14, 4.30, 0.06; xylose 54.79, 23.77, 4.86, 0.09; Klason lignin 47.73, 34.04, 3.31, 0.07; sex 68.90, 2.53, 10.52, 0.15; average stem weight 1.88, 64.98, 7.60, 4.04.
• Beleggia and colleagues (2023): year took 67.0 to 88.5 per cent of the variation in phenolic compounds, triterpenes, tocopherols and phytosterols, and genotype by year took 80.1 per cent of DPPH radical scavenging activity.

Images reserved in the article

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

• File:Cannabis-G-by-E-variance-shares-by-trait.jpg (Measuring and partitioning), a chart to be drawn: stacked bars of genotype, environment, interaction and remainder for grain yield, biomass, plant height, days to maturity, total THC and total CBD in Campbell and colleagues (2019), and seed yield, biomass, plant height and emergence to full flowering in Ferfuia and colleagues (2021).