Alternative Approaches to Loblolly Pine Breeding Value Predictions

The objective of this research is to assess the relative contribution of separate covariance matrices to predict full-sib progeny performance from crosses among a small number of parents, by modeling gene structure and gene expression in cross-validation studies.  Questions we intend to address include: (1) can we obtain reasonably reproducible results from triplicate samples of seedlings from OP, PMX, or CP families with respect to estimating family-mean levels of gene expression for a set of parents; (2) can we identify methods for combining those family-mean estimates of gene expression levels into covariance estimates for pairwise-combinations of parents that show utility in cross-validation studies for modeling phenotypic variation, and (3) do covariance matrices based on coding sequence SNP variation, gene expression level variation, or pedigree-based estimates of allele sharing have independent value for modeling phenotypic variation, or are they redundant so that one approach has the same information present in the other two?  To answer these questions, we have completed RNA-seq on 43 different families using biological replication of pools of whole seedlings harvested by family at age 3-months.  Family-mean gene expression patterns were used with phenotypes from age 6 progeny tests to identify genes or clusters of genes deemed as having a relationship with the trait.  By utilizing covariance matrices of these genes we have produced cross validation predictions with R² values of .5-.59.  To assess bias in our approach, we are sequencing an additional set of 31 families for further analysis.