July 9th, 2020
CIAT EA Beans
The number of parents, crosses and progeny per cross can have an impact in the rate of gain in the short, medium and long term. When too big for a fixed period, the programs can have lower rates per dollar invested.
Crossing
\(\sigma\), i
Finding the optimal number of parents, crosses and progeny will maximize rate of gain per dollar invested.
| Treatment | Description |
|---|---|
| 40-100-180 | Current scheme, using 40 new parents in the crossing block (20 at PYT, 5 at VEF, and 10 at OF), making about 100 crosses and generating 180 progeny per cross. |
| GRID | A grid of #of parents (10-60), crosses (45-190), progeny per cross (100-400) constraining from 2000 to 18000 generated individuals entering the SSD pipeline. |
A 20 year burn-in period was modeled using the current breeding scheme. The burn-in was followed by a 50 year evaluation period to measure rates of genetic gain for all treatments. Genetic gain was measured by assessing changes in genetic merit in the F2 population. Genotype-by-year interaction variance was assumed to be equivalent to genetic variance (based on average correlation between locations being equal to 0.5). All evaluations were conducted using 30 replications.
We simulated 4 complex and 3 simple traits to be behind the genetic merit and inferred through a selection index.
By year 50, the best crossing strategy (small #of parents) produces 1.08 (95% CI: 0.95,1.24) times more gain, than the current practice (40 parents).
Cassava results show that for a fixed time period, if you were able to do all possible crosses for different number of parents, there’s an optimal number of parents and crosses. Intensity plays a role.
If we do not assume we have the optimal size what simulations show is gain is asymptotic (there’s no maximum but there’s a diminishing gain). Optimal program is the smallest one (always diminishing).
If we assume we have all universe of possibilities, we can assume a distribution and pick the mean as the optimal size that maximizes gain per dollar invested.
The simulations showed that smaller Ne can increase the rate of gain up to ~10% compared to programs having greater Ne (constraining the plot number and a fixed time period). We recommend reducing slightly the number of parents (i.e. 20), and using a convenient #of crosses and progeny (i.e. 20-80-220) to maximize the genetic gain in the medium term (~50 years) given the current selection scheme.