Total Merit Index - the Breeding Goal for Swedish Dairy Cattle

Successful breeding requires that the breeding goal is properly defined. This is usually done by defining a total merit index (TMI) to be used as the main selection criterion in the population. This report will focus on the definition of the Swedish breeding goal. The first requisite for a goal that includes functional traits is the availability of relevant information. The first part of the report describes how the collection is organised and the history behind it. It is followed by a description of the time for the inclusion of different sub-indices and a broad presentation of how these individually are calculated. Finally the construction of the Total Merit Index and its expected effects are described.

Milk-recording and AI data in integrated data bases

Properly collected and processed field data is the corner stone in an efficient selection programme for dairy cattle. In the sixties AI and milk recording were reorganised and integrated in Sweden . EMIL was the name of the first computer, which was bought in 1961 in order to create an integrated processing of data from both these sources. In this respect Sweden became a pioneer. Records of inseminations and calvings became available in the same computer and the pedigrees of all animals were automatically updated (see graph). Many dominating countries in dairy cattle breeding have been lacking this instrument until quite recently and consequently, the possibility to select for female fertility, calving difficulties and stillbirths, based on all milk recorded cows. A high proportion, 86%, of Swedish dairy cows is connected to the official milk recording.

The time for inclusion of different sub-indices

When the first TMI was introduced in Sweden in 1975 it included growth rate, daughter fertility, stillbirths, conformation traits and temperament parallel to yield (Philipsson et al., 1975). For long the ambition was to include disease resistance in the breeding goal based on the use of the veterinary diagnoses. This was achieved in 1986, when official proofs were calculated for mastitis resistance and resistance against “other diseases” for the first time. The last contribution to the list of traits in the TMI is stayability, which was added in 1995.

Many of the traits mentioned have low heritabilities. The international acceptance of low heritability traits in the breeding goal for dairy cattle was delayed by the well known fact that mass selection has little effect on these traits. Mass selection is based on the performance of each of the individuals considered. What the critics failed to realise was that dairy cattle breeding is not just mass selection. Progeny testing of bulls is the main component. Selection

by means of bull proofs based on more than 100 daughters is worthwhile even if the heritability is below 5 per cent.

When a new trait is added to earlier ones, it often receives an observation status during the first years, and it is considered for selection only after this period. In Sweden all the traits mentioned have passed these research and observation phases and they are all weighted together in the TMI.

The design of the main subindices

All Swedish subindices are given a mean of 100 (mean for the young bulls of the last three years) and have a standardised genetic standard deviation of 7 units (Eriksson, 1999).

The general design of the main subindices is as follows:

The Milk index. The trait is based on the three indices: kg milk, kg protein and kg fat. The b-values of these indices are –0.10, 1.00 and 0.10 respectively for SRB, and –0.20, 1.05 and 0.20 for SLB. The negative weights on kg milk means that the milk index will promote an increase of the concentration of the milk constituents.

Daughter fertility. The trait is based on the following indices:

• Number of inseminations of virgin heifers and cows in the first and the second lactation
• Heat score (subjectively estimated) of virgin heifers and cows in the first and the second lactation
• CFI, number of days from calving to first insemination of cows in the first and the second lactation
• Number of fertility treatments in the first lactation.

Calving index. This index is calculated both as a trait of the sire of the calf and as a trait of the maternal grandsire. Both are based on indices referring to the frequency of stillborn calves and the course of the calvings.

Mastitis resistance. The trait is based on two indices, a diagnose of clinical mastitis or not, and the cell count in the period 10 days before to 150 days after the first calving.

Stayability. The trait is based on the proportion of daughters surviving the second lactation. A corrected stayability index is included in the TMI. It is corrected for the effect of yield, daughter fertility, mastitis and other diseases for SRB and in addition for the calving indices for SLB. The correction is performed in order to avoid double counting of the subindices for the traits mentioned, as they are correlated with (affecting) stayability.

The Total Merit Index, TMI

Selection index theory (Hazel, 1943) is applied at the construction of the TMI. This requires that heritabilities and genetic correlations are estimated. Actual estimates among the traits in the present TMI are given in Table 1 together with the heritabilities used.

Table 1. Approximated genetic correlations for SRB och SLB according to Calo et al. (1973), SRB above the diagonal. SLB below the diagonal. On the diagonal, the heritabilities used. Empty squares have the value of zero.
	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.
1. Milk index	.24	.19	-.24			-.23	-.19
2. Beef index	.13	.34	.15	-.27	-.21	-.15		-.13		.26	.43
3. Daught. Fertility	-.36		.055		.27	.25	.42
4. Calving indexS		-.16		.042	.46				.18	-.38
5. Calving indexMGS			.37	.55	.042	0.25			.23		-.18
6. Mastitis	-.26				.24	.035	.38	.30
7. Other diseases	-.23		.42		.28	.46	.03
8. Udder conform.	-.18		-.20			.63	.36	.20	.15	.26
9. Leg conform.						.22	.25	.34	.20
10. Height withers										.25
11. Tempe- rament											.12
12. Staya- bility												.03

The “normal” procedure at the construction of a selection index is to calculate the economic weights, which tells how much the increment of one unit of each subindex means for the net profit of a cow in a year. It should be noted that all desired effects cannot be measured directly in monetary terms. Ethical and other considerations may as well be included in the economic weights, e.g. for calving performance. The input parameters are besides the economic weights, the heritabilities of the different subindices and the genetic correlations among them as given in Table 1 and in addition the phenotypic correlations and the correlations between the measured traits and those in the breeding goal.

The most important output parameters are the so called b-values. The formula for a TMI is written as follows:

TMI = b 1 (X 1 -100) + b 2 (X 2 -100) +…..

The b-values are partial coefficients of regression and tell how much the TMI is changing when the corresponding subindex changes by one unit. The Hazel-procedure finds the b i- values that maximise the genetic gain in economic terms. Table 1 illustrates how the TMI is calculated for an actual SLB (Holstein) bull in Sweden by means of the b-values.

Table 2. Total merit index for the SLB bull 45013 Gubbilt
Trait	Relative breeding value	b-value, weight of deviation from 100	Contribution to the total merit index
Production
Milk index Beef index	120 104	1.00 0.20	20.00 0.80
Fitness traits
Daughter fertility Calving index, sire effect Calving index, MGS effect Mastitis resistance Other diseases Temperament Residual stayability	90 104 100 103 106 104 102	0.35 0.10 0.30 0.40 0.10 0.10 0.20	- 3.50 0.40 0 1.20 0.60 0.40 0.40
Conformation
Udder composite Legs	105 98	0.40 0.30	2.00 -0.60
Total merit index			21.7

The b-values used for SRB-bulls deviates from the ones shown in table 1 only for calving index MGS, where the SRB-value is 0.20 and for legs, where the SRB-value is 0.2.

When the first TMI was designed in 1975 nothing was known about the actual genetic correlations, so they were assumed to be zero. Lindhé and Philipsson (1998) showed that many of these correlations were significantly different from zero as Table 1 illustrates. This has two consequences. The first one is that the b-values are no longer identical with the economic values. The second and major consequence is that the expected genetic gains deviate considerably in comparison with a situation where the correlations really are zero. The gain in a subindex can be negative although the b-value is positive. The negative genetic correlation between yield and daughter fertility means, according to the theory, that selection for yield without consideration of daughter fertility results in a deterioration of daughter fertility, which has been observed in the Holstein populations in different countries (Philipsson et al., 1996; Philipsson and Lindhé, 2003).

Expected genetic gains

In popular science articles, b-values are sometimes presented as proportions. However, the proportions do not illustrate the distribution of the genetic gain expressed in monetary units, which calculations used in the selection index procedure will do. In Table 2 the results are given for SRB and SLB. The genetic correlations given in Table 1 have been used at the calculations and for both breeds it has been assumed that a modern AI selection programme is used in a population comprising 600 000 cows. The table also illustrates how the expected genetic gain, expressed in monetary units is distributed on the different traits.

Table 3. Expected future genetic gains for Nordic populations of SRB and SLB of equal size
Trait	SRB Index units Per cent per year economic gain	SLB Index units Per cent per year economic gain
Milk index	1.55 66.9	1.38 64.9
Beef index	0.61 3.6	0.56 4.0
Daughter fertility	0.29 4.5	-0.12 -2.6
Calving indexS	0.13 0.7	0.27 1.3
Calving index MGS	0.48 1.4	0.72 4.1
Mastitis	0.35 6.0	0.56 11.4
Other diseases	0.12 0	0.49 -2.4
Udder conformation	0.71 8.9	0.61 9.5
Leg conformation	0.44 1.8	0.73 7.7
Withers height	0.33 3.2	- -
Temperament	0.23 1.0	0.12 0.5
Corrected stayability	0.20 1.6	0.20 1.6

It is quite obvious that the genetic gain we can expect from the selection now applied is comprised by the gain in yield to about two thirds, whereas one third depends on other traits.

The differences between the breeds in genetic gain, which is most pronounced for the milk index and for the daughter fertility index, is a reflection of the different genetic correlations. The expected genetic gain in daughter fertility for SLB is negative. It is thus a challenge for the decision makers to achieve a positive trend, considering that most of the sires of sons are chosen from other countries.

Realised genetic trends, economic weights and b-values need to be continuously monitored to make sure that positive trends are achieved in the various traits according to the adopted breeding objectives

Cited literature

Calo, L.L., McDowell, R.E., VanVleck L.D. and Miller P.D. 1973. Genetic aspects of beef production among Holstein-friesians pedigree selected for milk production. Journal of Animal Science. Vol. 37, no. 3, 676.

Eriksson, J.-Å. 1999. Nytt Avelsmål. Swedish Dairy Association. FJE99034.

Hazel, L.N. 1943. The genetic basis for constructing selection indexes. Genetics 28:476.

Lindhé, B. and Philipsson, J. 1998. Genetic correlations between production with disease resistance and fertility in dairy cattle and consequences for total merit selection.Acta Agric. Scand., Sect. A, Animal Science. 48: 216-221.

Philipsson, J., Janson,L. and Brännäng, E. 1975. Selection index of bulls regarding economically important characters ( Eng. summary). Report Agric. College , Ser. A, No. 238, Uppsala , Sweden .

Philipsson, J., Banos, G. and Arnason, T. 1994. Present and future uses of selection index methodology in dairy cattle. J. Dairy Sci. 77:3252-3262.

Philipsson, J. and Lindhé, B. 2003. Experiences of including reproduction and health traits in Scandinavian dairy cattle breeding programmes. Livestock Production Science 83 (2003) 99-112.