Line by Tester in rice: GCA of lines, SCA of crosses, parent shortlist
Eight rice lines crossed to three testers for grain yield. GCA of lines and testers, SCA of the 24 crosses, and which parents and crosses move into the next stage.
Meera works on hybrid rice. She has eight promising lines (L1 to L8) from her own programme and three established testers (T1, T2, T3) with known general combining ability. She crosses every line with every tester, giving 24 crosses, and evaluates the crosses for grain yield per plant in an RBD with three replications. This is a Line by Tester mating design.
Question
Which lines have the best general combining ability for yield, which specific cross combinations have the highest specific combining ability, and is the trait driven more by additive or by non-additive gene action?
Data, in StatVeda format
Cross means matrix. One row per line, columns are the three tester means in order T1, T2, T3. Grain yield per plant in grams.
# rows = L1..L8, columns = T1, T2, T3 24.1, 26.8, 25.2 27.5, 30.1, 28.4 22.3, 24.0, 23.1 29.8, 32.6, 31.0 25.0, 27.2, 26.1 23.4, 25.1, 24.0 28.2, 30.9, 29.5 26.1, 28.4, 27.2
What she does in StatVeda
Open Plant Breeding. Pick Line x Tester Analysis. Set eight line names and three tester names. Paste the cross-means matrix. Run.
ANOVA: Crosses (23 df) highly significant. Partitioned into Lines (7 df, MS large, p < 0.001), Testers (2 df, p < 0.001) and Lines x Testers (14 df, p = 0.004).
GCA of lines (deviation from grand mean of 27.0 g): L4 +4.1 (best), L2 +1.9, L7 +1.6, L8 +0.2, L5 -0.7, L1 -1.6, L6 -2.5, L3 -3.9 (poorest).
GCA of testers: T2 +1.4 (best), T3 +0.1, T1 -1.5.
SCA of crosses: the largest positive SCA effects are L4 x T2 and L7 x T2. The cross L4 x T2 also has the highest per-se mean (32.6 g).
Variance components: sigma squared GCA = 4.82, sigma squared SCA = 0.41. Proportional contribution: lines 71 percent, testers 18 percent, line x tester 11 percent of the cross sum of squares. Additive (GCA) variance is about twelve times the SCA variance.
What it means
Yield here is overwhelmingly additive. The lines explain most of the variation among crosses and sigma squared GCA dwarfs sigma squared SCA, so a parent with good GCA reliably produces good crosses. L4 is the standout general combiner, L2 and L7 are solid, and L3 and L6 drag every cross they enter. On the tester side, T2 is the best general combiner. The significant but small line x tester term means a few specific combinations beat their GCA prediction, but the effect is not large enough to override the GCA-driven ranking.
Retain L4, L2 and L7 as parents for the next round of crossing: all three have strong positive GCA.
Drop L3 and L6 (large negative GCA, poor in every combination).
Advance the cross L4 x T2 to multi-location yield testing: it tops both per-se mean and SCA, with both parents being good general combiners, which is the most reliable kind of high cross.
Because additive variance dominates, recurrent selection on GCA is the right longer-term strategy, not chasing individual high-SCA crosses.
Meera writes up the Line by Tester section with the GCA tables, the proportional contribution, and the L4 x T2 recommendation. The clean dominance of GCA variance is the single number that decides the breeding strategy.
Why Line by Tester and not a full diallel
A full diallel among the eight lines would mean 28 crosses just among the lines, and it answers a different question (combining ability within one set of parents). Line by Tester deliberately uses a small number of common testers to rank a larger set of new lines economically. With three testers she gets a defensible GCA estimate for every line at a fraction of the crossing effort, which is the standard early-stage hybrid screening tool.
What she will do next season
Promote L4, L2, L7 and the L4 x T2 cross. Run a multi-location trial of the top crosses next season and then read GxE structure with an AMMI or GGE biplot. Line by Tester filtered the parents; the multi-location stage measures whether the leading cross holds up across environments.