INTRODUCTION TO PLANT BREEDING

AGRONOMY 815 / COURSE NOTES

P. STEPHEN BAENZIGER, 338 Keim Hall, 472-1538

DEPARTMENT OF AGRONOMY / UNIVERSITY OF NEBRASKA

THE EFFECT OF SELECTION ON QUANTITATIVE TRAITS

Fehr, Chapt. 17 — very complicated, use as a secondary reference
Briggs & Knowles, p. 196-198.
Allard, Chapt 10.
Simmonds, p. 97-110.
Falconer, Chapt. 11.

Selection with a well-expressed major genes is simple; with polygenes it is not simple because they show small individual effects and relatively large environmental ones. For any individual locus, s (fitness) is usually very near 0. Also, we cannot determine the effect of selection at a locus (this may change with RFLPs or other genetic markers).

If S is the mean phenotypic value of the selected parents, expressed as a deviation from the population mean, and R is the mean deviation of the their offspring then S is the selection differential and R is the response. They are related by:

Note: — If you are selecting among homozygous individuals ( i.e., in a land race) then would use

H2 (broad sense heritability).

One of the main uses or needs for determining the heritability of a character is that it can be used in prediction equations. The prediction is good only for one generation in that selection will change the gene frequencies and will reduce variability, hence, heritability (however, generally the changes in heritability are small). For the prediction, can use previously determined values for heritability for the population. Though not technically correct because heretability is unique to a population, we can obtain an idea of heretability from other "similar" populations.

Example: Have a h2 = .61. Mean of the total population for yield is 40 bu/a. The selected parents have a mean of 50 bu/a. Hence, the selection differential is 10 bu/a.

R = (.61)(10) = 6.1 bu/a. The resulting population should have a mean of 40 + 6.1 = 46.1 bu/a.

If the phenotypic values are distributed as a normal curve and selection is by truncation then:

S = is p where i is the intensity of selection.

i is a property of the normal curve and is determined only by the percentage that are selected. For values of i between 0.1 and 0.001, i = 1.13 + 0.731 log (1/k) where k is the proportion of the population that is selected. This approximation assumes at least 20 individuals in the unselected population.

RAMIFICATIONS:

  1. R = ih2sp = ihs a.

  2. By decreasing the number of plants selected from 10% to 1%, the intensity of selection is only increased from 1.86 to 2.59. Hence there is a diminishing return by increasing the selection pressure.

  3. 3. Best response occurs when sp is reduced, the mean of the parents is high, and to some extent when the selection intensity is increased.

Note — can obtain an estimate of heritability which is called realized heritability by:

R/S = h2. Difficulty is that realized heritability is often confounded and does not accurately estimate heritability.

Type of empirical responses to selection and the interpretation of results.

  1. Respond approaches a plateau  expected result as the variation within the population is reduced and the gene frequencies of favorable alleles are fixed.

  2. Response continues without a plateau for a long time  large variation in the original population or new variation is identified by continued breaking up of linkage groups.

    Example Illinois High (and Low) Protein and Oil Corn Populations.


  3. Response reaches a plateau and then suddenly begins to increase again  an unfavorable linkage was broken.

  4. Isolated subpopulations at a plateau are intermated with an increased response and a move to a new, higher plateau  removal of the random fixation of deleterious genes by intermating.

  5. Apparently useful variation exists at the plateau stage  heterozygotes have an advantage.

FAMILY SELECTION — superior progenies are identified on the basis of replicated field trials preferably grown in more than one environment.

3 phases in family selection:

  1. FORMING FAMILIES;

  2. EVALUATING FAMILIES AND SELECTING THOSE THAT ARE SUPERIOR;

  3. INTERCROSSING PLANTS PRODUCED FROM REMNANT SEED OF THE SELECTED FAMILIES (OR SELFED SEED OF PARENTS) TO FORM THE IMPROVED BREEDING POPULATION FOR THE NEXT CYCLE OF IMPROVEMENT.

HALF-SIB FAMILY TESTING AND SELECTION

EAR-TO-ROW method in corn . . . select good ear and plant in row harvest best ear from best rows.

Method:1.Females pollinated with a random sample of pollen from the same population to produce seed of HS families.

2.HS families tested in trials.

3.Selected families recombined (intercrossed).

A modification of this method was proposed by Lonnquist in 1964. This involves selection within HS families as well as among HS's  combined selection. Shown to be very effective in improving yield in corn. (Webel & Lonnquist. 1967. Compton & Bahadur. 1977) This method has been modified further (Compton & Comstock. 1976).

FULL-SIB FAMILY TESTING AND SELECTION

Method:1.Crosses are made between pairs of plants.

2.FS families tested in replicated trials.

3.Selected families are recombined (intercrossed).

FS selection requires only 2 generations per cycle when plant to plant crosses are made between plants from different selected families because recombination and family formation are accomplished in one operation. Note not as complete a recombination cycle. (Moll & Robinson. 1967. Der Zuchter; Moll & Stuber. 1971. Crop Sci.)

S1FAMILY TESTING AND SELECTION

Method:1.Random sample of plants in the base population are selfed.

2.Selfed progenies (S1) are evaluated in replicated trials. Know a lot about parents = progeny test.

3.Superior S1 families are recombined.

This method has not only proved very effective in improving performance in cross pollinated crops, but it has increasingly been utilized in naturally self-pollinated species. It tends to give great within and among family differences. Use of selfing makes it an excellent approach for improving self-pollinated crops. e.g., sorghum ( Ross.1978. 33rd Ann. Corn & Sorghum Res. Conf.)

S2FAMILY TESTING AND SELECTION — generally used as part of an inbred line development program.

Method:1.S2 families are formed by selfing S1 plants.

2.2 families evaluated in replicated trials.

3.Superior S2 families are recombined.

Greater gains per cycle should be possible but an extra generation is necessary. (Horner et al., 1973) O.K. if the extra generation can be done in an environment between selection environments.

S1FAMILY SELECTION BASED ON TESTCROSS PERFORMANCE

Method:1.Individual plants are selfed and also crossed to a tester:

· several random plants from a population;
· single cross hybrid;
· inbred line.

CHOICE OF A TESTER IS CRITICAL

2.Testcrosses (topcrosses) evaluated in trials.

3.Selfed seed corresponding to the superior testcross combinations is grown and these S1 recombined.




INTERPOPULATION IMPROVEMENT

Rationale . . . Breeding systems based upon additive genetic variance not likely to be the most efficient where the trait for which selection is practiced exhibits heterosis. In such cases heritability tends to be low and the non-additive genetic variances may be of considerable magnitude.

A system designed to make maximum use of all types of genetic variance was proposed by Comstock et al., 1949. Agron. J. 41. Reciprocal Recurrent Selection procedures are all based on this original scheme of Comstock, Robinson and Harvey's.

BASIC PROCEDURE:

  1. Plant two different populations. In each population self several selected individuals and also cross them to the other population (several random plants HS; paired plant-to-plant FS; inbred line from other population). Basically each population becomes the tester for the other population.

  2. Testcross progenies evaluated in replicated trials. Individuals with superior combining ability are identified.

  3. Selfed seed of each superior original plant is grown, keeping populations separate. Intercross within each population. GCA/SCA





GSA/SCA

Two important terms that describe how the parent lines of a hybrid interact are General Combining Ability (GCA, also known as average combining ability) and Specific Combining Ability (SCA). General combining ability refers to how a line, in general, "combines" (forms hybrids) with other lines as measured for a trait. General combining ability is often considered as an indicator of additive gene action. Specific combining ability refers to how a line specifically combines with another line and includes both additive gene and dominant gene action. One way to visualize the differences between GCA and SCA is to consider a series of crosses among lines (inbred or populations). The crosses can be analyzed as a factorial design where the effect of individual lines is the main effect (estimates GCA) and the interactions of the individual lines (equivalent to the interactions of the main effects) estimates SCA. The statistical model would

be:

Trait = m + Ri + Mj + Fk + (M x F)jk + error

where:m is the mean
R is the replication effect
M is the effect of male lines (part of the estimate of GCA)
F is the effect of female lines (part of the estimate of GCA)
M x F is the interaction of males and female lines (the estimate of SCA)

The ANOVA for diallel crosses (n(n-1)/2 crosses) of n parents is often written as follows:

Source

df

Replications

Crosses
GCA
SCA

Error

r -1

[n(n-1)/2] -1
n-1
n(n-3)/2

(r-1){n(n-1)/2-1}

Total

rn-1