BeesCene Volume 23, #4, DECEMBER 2007

Selection Indices 101: Selection for Multiple Traits

by Adony Melathopoulos

As queen breeders we know that our customers are not interested in stocks that have been selected for a single trait, rather they expect that our stocks to be improved across multiple traits. Selection on multiple traits, however, is not as straight forward as selecting for a single trait. I know of three ways to accomplish this kind of selection (tandem selection, selection based on independent culling levels and selection using an index). In this article I will try to convince you the best strategy is the use of a selection index, I will explain how the BC Bee Breeders’ project is thinking of making its selection based on an index and explain the survey on pages 29-30, which help shape the program’s own index.

There are three ways to select for multiple traits. The first is tandem selection, which involves selection for one trait at a time, such that selection for a first trait is followed by selection for a second trait, third trait, etc., until the desired level of improvement is reached. The second strategy is selection based on independent culling levels, which involves the establishment of a level of merit for each trait and culling any breeder that does not meet this level. So, for example, you might cull any colony that does not make at least 100 lbs of honey, chased you out of the yard, and showed any sign of chalkbrood. While I suspect that most of us have used these two techniques in the past, years of research across numerous livestock species have shown that both tandem selection and independent culling levels are vastly inferior to selection by index.

The idea of a selection index was developed in the 1940s by the livestock breeding wizard Lanoy Nelson Hazel. Hazel had spent several summers in California, working on problems of breeding for improvement of many different traits of poultry, including age at sexual maturity, rate of egg production, egg size, shell strength, internal egg quality, and viability under differing exposures to disease vectors. Although he realised that breeders for centuries had been selecting on multiple traits, there was a need for a systematic method for making the selection. His mentor Jay Lush acknowledged that some of this problem was entirely psychological:
What is involved here isn’t the genetics of the case but human psychology in being able to keep two or more things in proper balance with each other without help from some such device as a score card or selection index.
The other part of the problem, however, is to incorporate two concrete factors into the decision: 1) the economic value of the trait and 2) how responsive the trait is to selection (a category that breeders call heritability).

I won’t give you anything but the broad contours of how to calculate a selection index here. If you really must know I suggest you pick up Dr. Ernesto Guzman’s new and inexpensive book on bee breeding (Elemental Genetics and Breeding for the Honeybee).

You begin by assessing a pool of colonies within your operation for a variety of economically important traits. These traits are invariably measured in incomparable units (ie honey is in lbs., hygienic behaviour is in percentage of brood removed in 48h, varroa resistance in the number of varroa that drop per day) and come from colonies living in very different apiary environments. To account for yard to yard differences and to convert all the measurements into a standard unit you convert all the measurements using some rudimentary statistics into units known as Z-scores. Z-scores essentially make it possible compare apples to oranges, or more specifically, traits measured using different units from colonies in different yard environments. The Z-scores for each measured trait are then multiplied by the heritability of the trait (if it is known) and the economic value of that trait, and then all these products add up to give you a selection index for that colony. The colonies with the highest index values are used as breeders.

If you were able to follow the gory details of this highly abridged description of how to calculate a selection index you will have noticed how critical the economic value of each trait is in driving selection decisions. This is the power of this method of selection. Traits that are only moderately valuable to your customers will be weighted far more lightly than traits they deem economically essential. Consequently, the selection process will more efficiently deliver what your customers want.

This is where the attached survey comes in. In order to determine the economic value we are using a survey developed at the University of Guelph that allows you to define the values for each trait. To accomplish this we want you to assign 100 points across number of traits. Think of it like this, what if queens cost $100 (heavens forbid)? How would you spread that $100 across the various traits? Would you spend $50 on honey production, $20 on varroa resistance and spread the remaining $30 evenly across the remaining traits? Maybe you think queen colour is pretty important, in which case you would fill the category titled “other” with “dark coloured bees” and put that remaining $30 there. Whatever you want, just make sure total comes out to $100.