Simple robust genomic prediction and outlier detection for a multi-environmental field trial
share
The aim of plant breeding trials is often to identify germplasms that are well adapt to target environments. These germplasms are identified through genomic prediction from the analysis of multi-environmental field trial (MET) using linear mixed models. The occurrence of outliers in MET are common and known to adversely impact accuracy of genomic prediction yet the detection of outliers, and subsequently its treatment, are often neglected. A number of reasons stand for this - complex data such as MET give rise to distinct levels of residuals and thus offers additional challenges of an outlier detection method and many linear mixed model software are ill-equipped for robust prediction. We present outlier detection methods using a holistic approach that borrows the strength across trials. We furthermore evaluate a simple robust genomic prediction that is applicable to any linear mixed model software. These are demonstrated using simulation based on two real bread wheat yield METs with a partially replicated design and an alpha lattice design.
READ FULL TEXT