Best linear prediction of breeding values in a forest tree improvement program

Area of Science:

• Forestry
• Quantitative Genetics
• Plant Breeding

Background:

• Traditional methods for predicting breeding values in forestry often use averaged standard scores.
• These methods may not fully leverage complex progeny test data.
• Best Linear Prediction (BLP) and Best Linear Unbiased Prediction (BLUP) are established in animal breeding but less common in forestry.

Purpose of the Study:

• To apply Best Linear Prediction (BLP) for predicting breeding values in a slash pine (Pinus elliottii) breeding program.
• To compare BLP rankings with traditional averaged standard score rankings.
• To assess the suitability of BLP/BLUP for handling complex, real-world progeny test data in tree improvement.

Main Methods:

• Utilized Best Linear Prediction (BLP) on progeny test data from 1,396 slash pine parents.
• Compared BLP-derived parent rankings against rankings based on averaged standard scores.
• Employed an approach treating fixed effects as known and accounting for heterogeneous variance structures across environments.

Main Results:

• BLP identified higher-ranking parents associated with a greater number of more precise progeny tests.
• Standard score rankings showed an inverse trend, favoring parents from fewer, less precise tests.
• The study adapted BLP/BLUP to handle inappropriate assumptions of homogeneous variances common in animal breeding applications.

Conclusions:

• BLP offers a more robust method for predicting breeding values in forestry compared to standard scores.
• BLP and BLUP can effectively manage complex and 'messy' progeny test data in tree improvement.
• These methods are valuable for increasingly sophisticated forest tree breeding programs seeking to incorporate diverse data sources.