Effects of systematic data reduction on trend estimation from German registration trials

Area of Science:

• Agricultural Science
• Plant Breeding
• Quantitative Genetics

Background:

• Variety trials evaluating a variety's value for cultivation and use (VCU) are crucial for assessing genetic progress in crops.
• VCU datasets include registered varieties and genotypes that did not achieve registration, forming distinct post-breeding and post-registration populations.
• Excluding data from genotypes not reaching registration can introduce bias into genetic trend estimations.

Purpose of the Study:

• To investigate the impact of data exclusion (informative drop-out) on genetic and non-genetic trend estimations in wheat breeding.
• To determine if VCU trial data, when fully utilized, provides unbiased estimates of post-breeding trends.
• To quantify the bias introduced by restricting data to only registered varieties or those tested for a minimum duration.

Main Methods:

• Analysis of historical German winter wheat VCU trial data.
• Simulation study designed to mimic the structure of VCU trials and assess the effect of data exclusion.
• Comparison of trend estimates derived from complete VCU datasets versus restricted datasets (e.g., genotypes tested for ≥3 years).

Main Results:

• When all VCU trial data was used, unbiased estimates of post-breeding trends were obtained, even when simulating zero true trends.
• Restricting data to genotypes tested for at least three years resulted in biased estimates: a positive genetic trend (+0.11 dt ha⁻¹ year⁻¹) and a negative non-genetic trend (-0.11 dt ha⁻¹ year⁻¹).
• The observed bias increased with higher genotype-by-year variance and was mitigated by random data selection.

Conclusions:

• VCU trials offer unbiased estimates of post-breeding trends when all available data is incorporated.
• Excluding data from genotypes tested for shorter periods (e.g., up to two years) can lead to significant biases in trend estimations.
• The findings highlight the importance of utilizing complete VCU datasets to accurately assess genetic gains and understand yield improvements in wheat breeding.