Selection indices for quality evaluation in wheat breeding

Theor Appl Genet (1992) 84:57-64 N 9 Springer-Verlag 1992 Selection indices for quality evaluation in wheat breeding G. Branlard 1,., j. Pierre 2, and M. Rousset 1 INRA, Station for the Improvement of Plants, 63039 Clermont-Ferrand, France 2 INRA, Station for the Improvement of Plants, 35650 Le Rheu, France Received August 5, 1991; Accepted November 11, 1991 Communicated by H. F. Linskens Summary. From multilocation trials involving 125 cultivars of wheat of mainly French and European origin four tests - protein content, Pelshenke, modified Zeleny and the mixograph - were used to establish six selection indices. Three of these indices - IW1, IW 2 and IW 3 - were calculated in order to evaluate the genetic potentiality of the lines for dough strength as given by the Chopin alveograph. The indices IV1, IV2 and IV3 were established to evaluate loaf volume as measured by the French bread-making standard. A quality index IQ was calculated from the allelic effects of the high-molecular-weight (HMW) subunits of glutenin from 195 cultivars assessed by the Chopin alveograph and the Pelshenke test. Comparison of the relative efficiency of each of the six indices to the individual tests revealed the superiority of the indices over one or several technological parameters. The six selection indices and the quality index were compared using 30 very diverse 174 lines. Their ability to retain the good quality lines is discussed in particular. Key words: Bread wheat - Technological quality - Loaf volume - Gluten strength - Heritability Introduction Breeders use various tests to evaluate wheat quality. These tests are based on the end-use quality objective of the breeding programmes, but there are also one or several indirect technological tests commonly used in a particular country and they must be feasible in the F3, F4 or F s generations. In addition to the analysis of the allelic composition of the high-molecular-weight subunits of glutenins (HMW-GS) (Payne et al. 1981) and of gliadins * Correspondence to: G. Branlard (Sozinov and Poperelya 1980) which can be carried out from single kernels, or half grains (Bietz et al. 1975), breeders may have many other technological tests at their disposal. Some tests ha~e been found to be more useful than others for predicting the quality of North American wheats (Baker et al. 1971 ; Fowler and de la Roche 1975). With respect to European bread wheats, Branlard et al. (1991) compared 17 technological tests for their ability to be both highly heritable and correlated to quality. Some technological parameters, for example, modified Zeleny, Pelshenke test and mixograph criteria, were more suitable for predicting gluten strength or loaf volume than others. As wheat quality is the result of many components, each having different technological properties, which cannot be evaluated by only one indirect test, it appears that a very judicious combination of a few test parameters is necessary for improving the efficiency of wheat quality breeding. Some studies have formed quality indices by combining several technological parameters (Cox et al. 1989). Surprisingly, no selection index has been established for wheat quality as has been for many other agronomic characteristics of plants. The theory of genetic indices first developed by Smith (1936) on wheat was used with success for improving the genetic value of different traits in many crops, such as grain yield in maize (Robinson et al. 1951) grain and straw yield in oats (Eagles and Frey 1974) or yield and protein content in maize (Motto and Perenzin 1982; see Baker 1986 for review). In the study presented here several indices will be established, six of which were calculated from a limited number of technological tests, each using a small amount of grain, and another from the allelic composition of the allelic composition of the HMW-GS. Their relative efficiency for improving the genotypic value of strength and loaf volume will be calculated and particularly discussed with particular reference to use F4 offspring. 58 consistency of the curve (MTM), height and thickness, respectively, of the curve at the maximum (MHM and MTH). Height and thickness, respectively, of the curve after 7 rain of mixing (MH7 and MT7), the product (MTM x MHM) and the sum of the six previously defined traits (MIN). The criteria for which the indices were developed are the strength W of the Chopin alveograph and the bread loaf volume LV as given by the French standard breadmaking procedure ( A F N O R V-03-713). Material and methods The six technological indices consisted of IW~, IW z and IW 3, established for improving the genotypic values of strength as evaluated by the Chopin alveograph, and IV1, IV2 and IV3 for bread loaf volume. Plant material The six indices IWi and IV~ were established from a total of 125 bread wheat cultivars of European origin and experimentally evaluated for 3 years (1985 1987) in six, seven and five locations, respectively, in France. These multilocation trials enabled us to compare 46 technological parameters (Branlard et al. 1991). Seventy other cultivars of very diverse genetic origins and range of quality were grown each year by the INRA wheat laboratory of Clermont-Ferrand. These seventy cultivars were used to relate gliadin and HMW-GS polymorphism to technological quality (Branlard and Dardevet 1985 a, b) and were added to the previous ones also grown at Clermont-Ferrand. The means of the technological values of these 195 cultivars were used to establish the quality index IQ from their allelic HMW-GS composition. The relative efficiency and the correlations between the indices IWi, IVi and IQ were computed fi'om the 125 cultivars. The indices were compared using 30 F 4 lines derived from offspring from several complex crossings including 8 diverse bread wheat cultivars as different combinations between 37 parents. This population was selected during a 3-year breeding programme from 1987 to 1989. In the second year, plants were selected for resistance to disease and for quality using IQ. In the third year these 30 lines were used as experimental material for agronomic purposes at seven locations, and their technological qualities were evaluated from only one location. Indices from technological traits The three strength indices, IW1, IW2 and IW3, and the three loaf volume indices, IV1, IV2 and IV3, were calculated by combining two or three of the following tests: Pro, Pel, Zym and Mixo. The mixograph test had seven parameters, consequently the maximum number of parameters used as explanatory variates xj of the strength and loaf volume was ten. The initial variates xj were standardized x'j = ( x j - m j ) / a j (where mj and aj are the mean and standard deviation of xj respectively), before calculating the multiple regression for predicting the strength W or the loaf volume LV. IW 1 and IV 1 were calculated as functions of three tests: Pro, Pel and Zym. IW2 and IV 2 were calculated from Pro, Pel and Mixo, whereas IW3 and IV3 used only Pro and Mixo. The optimum number p of explanatory variates x'j introduced in the multiple regression was obtained as previously described (Branlard and Dardevet 1985 a). The general formula of indices IWi and IVi is written as follows IVi o r l W i = K ~ ci f l j h 2 x J (1) j=l where h~ is the general heritability calculated for each parameter xi, as described by Branlard et al. (1991), flj is the partial regression coefficient of the step-wise multiple regression calculated to predict W or LV from the variates x], cj -- 1/xmj where xmj is the maximum value of x j, and K is a coefficient applied to the sum in order to obtain an index value near 100 when all xj values are equal to their maximum x m j. The h~, flj, xmj coefficients are given in Table 1. As the product Kcj flj h~ is constant, let be aj = Kcj flj h~ for a given j; then the index formula (1) can be written: Technological tests From the 17 technological tests previously compared, the following were retained to build the indices of technology IWi and IVi: protein content (Pro), Pelshenke (Pel), modified Zeleny (Zym) and mixograph (Mixo). The protein content was estimated on wholemeal flour by Near Infrared Reflectance (NIR). The modification of the Zeleny consisted in performing the test from a 70% instead a 20% extraction rate flour. From the 10 g mixograph, seven parameters were obtained: time taken to reach the maximum P IWi = Z aj xj j=l Table 1. Coefficients used in calculating the selection indices: h 2, general heritability; W i and Vi, flj coefficients obtained from the multiple regression to explain dough strength W and loaf volume LV, respectively Parameter Abbreviation Units Maximum Hz Wl W2 W3 Protein Pelshenke Zeleny modified Mixographe time at maximum Height at maximum Thickness at maximum Height at 7 min Thickness at 7 min MTM x MHM Index mixo Pro Pel Zym MTM %db rain ml rain 18 305 59 5.7 0.247 0.633 0.632 0.472 0.175 0.337 0.526 0.192 0.299 0.255 MHM MTH cm cm 9.4 5.6 0.437 0.403 MH7 MT7 MTM x M H M MIN cm cm - 7.9 3.2 36.0 47.5 0.546 0.565 0.557 0.618 Rz n = 125 VI V2 V3 -0.159 0.095 0.652 -0.123 0.145 0.091 -0.250 1.692 -1.574 0.365 0.333 0.728 0.316 -2.241 0.193 2.090 -0.233 0.81 0.78 0.60 0.60 -0.104 0.75 0.43 59 Quality index IQ from HMW-GS The quality indice IQ was calculated as the sum of the coefficients attributed to the HMW-GS alleles present in a given genotype. The coefficient attributed to each allele was globally proportional to the allelic effect observed for the following parameters: strength, tenacity, swelling given by the alveograph and Pelshenke. For a quality parameter k the effects were calculated for each locus and expressed in percent of the total variation Axp observed amongst the three loci Glu-A1, Glu-B1 and Glu-D1. The effect Xkij of the allele i from the locus j was obtained as follows: Xkij = ( X k i j - - Xkrj) X 100/Axk; where Xki j is the mean value of the parameter k calculated from the varieties having the allele i at the locus j; for example, mean of the swelling of the wheats having the band 2* at GIu-A1. Xkrj is the mean value of the parameter k calculated from the varieties having the allele inducing a low quality. For a given locus j the same allele r (named here in the reference allele) was used for the four quality parameters. This reference allele was Glu-Alc (null allele), Glu-Bld (bands 6 - 8 ) and GIu-DIa (bands 2-12) according to the Payne and Lawrence (1983) nomenclature. Axk was the difference between the maximal and minimal means. The mean effect X.i~ was calculated as the mean of the four effects Xk~j . The coefficient of quality Yij attributed to the allele i from the locus j was expressed as Yij = bR.ij + aj where b is a coefficient and aj is a constant corresponding to the quality value attributed to the reference allele. For Glu-Alc (null allele) a~ = 0 and for Glu-Bld and Glu-Dla, a 2 = 2 and a 3 = 7 , respectively. Finally, the indice IQ is written as the sum of the three Y~j coefficients: 3 IQ = 2 Yij j=, Comparison of the indices The values of the indices IW~, IV~ and IQ calculated for the 125 cultivars allowed us to compare them by two approaches. First- Table 2. Indices of selection for improving the genetic value of dough strength (IW1, IW2, IW3) and loaf volume (IV1, IV2, IV3). The units and abbreviations of the parameters are given in Table 1 IW 1 = IW 2 = 0.05 Pro + 0.25Pel 0.07 P r o + 0.31Pel IW 3 = 1.65 IV1 = - 0.09 IVz = - 0.12 IV3 = 0.88 + 0.39Zym - 0.08MTM+0.33MH7 + 0.31 MT7 Pro- 2.86MHM+I0.41MH7 +4.67MT7 Pro + 0.14 Pel + 0.98 Zym Pro + 0.36 Pel + 2.94 M H M - 4.87 MH7 + 0.43 MT7 Pro-27.15MHM+45.05MH7 -5.19MT7 ly, linear correlations were calculated for each pair of indices. Secondly, the indices were compared to the technological tests by computing their relative efficiency RI/Rt (Baker 1986); RI was the expected response to selection based on index I, and Rt the expected response to selection based on the test t alone, from which the index I was built. On the assumption that the two types of selection (with the index or with test t alone) used the same selection intensity it has been demonstrated that the relative efficiency can be calculated as follows: R! Rt r (Gt, I) ht (Baker 1986) where r(Gt, I) is the correlation between the genotypic values G t of the cultivars evaluated by the test t and the values of the indice I. The divider ht is the square root of the heritability h 2 of test t. Criteria of selection Thirty F4 lines were observed using either the individual parameters or each of the indices IWi, IVi or IQ. The lines having a value higher than M + SE were considered to be selected. M and SE are the mean and standard error of the 30 lines, respectively. Results Indices from technological traits T h e coefficients X m j , flj a n d h~ used to build the indices are s h o w n in Table 1. T h e m a x i m a l value o f each t e c h n o logical p a r a m e t e r was n o t always f r o m the s a m e cultivars. As expected, the heritability o f P r o was l o w ; those o f Pel, Z y m a n d M i n were higher a n d similar to one a n o t h e r . M u l t i p l e regression gave a higher R 2 coefficient for the p r e d i c t i o n o f the W v a l u e o f d o u g h t h a n for b r e a d l o a f v o l u m e LV. T h e lowest R 2 coefficient o f p r e d i c t i o n o f l o a f v o l u m e was o b t a i n e d for V1, i.e. w h e n the three tests Pro, Pel a n d Z y m were c o m b i n e d . S o m e p a r a m e t e r s o f the m i x o g r a p h p e r m i t t e d a significant increase in the p r e d i c t i o n coefficient for l o a f volume. T h e use o f the m i x o g r a p h , t o g e t h e r w i t h the tests P r o and Pel gave a high coefficient o f p r e d i c t i o n for strength ( R 2 = 0 . 8 1 ; n = 1 2 5 ) . T h e c o m b i n a t i o n s o f m i x o g r a p h p a r a m e t e r s were n o t the s a m e for the prediction o f W (Table 1), but p a r a m e t e r s M H 7 and M T 7 were regularly retained by the step-wise regression p r o c e d u r e in the p r e d i c t i o n o f W and LV. T h e six selection indices established f r o m these different coefficients are p r e s e n t e d in Table 2. F o r these indices Table 3. Means and standard deviations of the selection indices value observed for five wheats, ranked from the very good (Magdalena) to the very bad (Clement), over 3 years in France IW~ Magdalena Hardi Capitole Talent Clement 62.6 37.3 41.3 20.6 13.0 IW~ _+ 3.2 _+ 3.2 _+ 3.0 __ 1.1 _+ t.0 60.0 29.0 36.6 15.0 8.0 Iw~ _+ 5.3 _+ 5.0 _+ 3.2 _+ 1.0 _+ 0.2 76.0 70.0 63.6 61.3 47.6 IV~ _ 7.0 _ 5.5 _+ 4.6 _+ 3.0 + 2.0 68.0 50.3 47.7 30.7 20.0 IV~ 4- 2.6 _+ 3.8 _+ 2.5 _+ 2.0 _ 2.0 55.0 19.7 30.0 5.3 1.7 IV3 ___6.0 _+ 5.5 _+ 4.3 +_ 0.5 _+ 0.29 92.7 92.3 82.3 77.0 50.0 _+ 6.5 _+ 5.8 _ 5.1 • 2.0 _ 1.7 60 the negative signs correspond to t h e / / j signs and do not mean an unfavourable effect on quality for the parameters. As the variates introduced are the true parameters without any transformation, the use o f these indices is very simple. F o r five different wheats ranked from very poor quality (Clement) to very strong gluten (Magdalena) the magnitudes of the values for the six selection indices were different (Table 3). The magnitude o f the index value range was higher when the indices were constructed from tests Pro, Pel and Mixo (e.g. for I W z and IVz) than they were for the other combinations. I W 3 and IV 3 had higher values with a lower range between the bad and the good quality wheats than the other indices. F o r a given variety, the higher the index values, the higher their variations from year to year. The efficiency o f the indices were calculated relative to the parameters used. This relative efficiency corre- Table 4. Relative efficiency RI/Rt values obtained when the expected response RI to selection using the index (IWi or IVi) is compared to the direct selection Rt based on one of the parameters introduced in the index Pro Pel Zym MTM MHM MH7 MT7 W LV IW~ IW2 IW3 61.3 124.7 93.3 54.3 125.7 88.9 81.8 90.6 87.8 - 103.7 79.9 116.6 111.8 . . . 118.1 129.8 106.4 106.0 82.5 - 92.9 - I~ I~ 51.5 125.5 . 51.5 76.0 86.6 60.5 sponds to the comparison between the expected response to selection based on the index and the direct response when only one parameter is used. F o r each of the six indices, at least one parameter was found to be advantageously introduced as their expected response was superior when combined in an index to the response if it would have been used for a direct selection (Table 4). Selection based on IW 1 or I W 2 would be expected to result in a 24% or 25 % greater increase in the mean genotypic value of the strength than would direct selection for strength using the Pelshenke test alone at the same selection intensity. Selection based on the index I W 3 = 1.65 P r o - 2 . 8 6 M H M + I 0 . 4 1 M H 7 + 4 . 6 7 M T 7 would be expected to give a mean genotypic strength value 3 - 2 9 % greater than would result from the direct use of the individual parameters. As the protein content is partly related to strength and to loaf volume and its coefficient attributed in the indice is generally low, the expected response o f this parameter seems to be better when used alone. Nevertheless, the introduction o f protein content with some mixograph parameters in the indices led to a better prediction o f the genotypic values o f the criteria W or LV to be improved. I~ Quality index from H M W - G S 51.0 - F r o m the 195 wheat cultivars the mean value of the genotypes having the same allele was calculated for each of the three alveograph parameters (strength, tenacity, swelling) and for the Pelshenke test. The comparison of these means was carried out for the main alleles at the loci Glu-A1, Glu-B1 and Glu-D1 (Table 5). F o r each parameter the effect xkij of the alleles was calculated and their means X i~ are presented. The coefficients Y~j pro- 78.2 112.6 103.5 104.0 For abbreviations, see Table 1 Table 5. Comparison of the means calculated on four quality parameters for the different HMW-GS alleles encountered on the 195 cultivars. Means with the same letter are not significantly different at P=0.95 Glu-A1 Glu-B1 Glu-D1 Tenacity (mm H20) 2* 75 a 1 69 b Null 64 b 17-18 77 a 7-9 71 b 7-8 70 b 6-8 68.5 b 13-16 66 bc 7 65 c 21 53.4 d 5-10 3-12 4-12 2-12 74 68 64 63 a b bc c Swelling (cm 3) 1 21.5 a 2* 21.4 a Null 20.1 b 13-16 20.9 a 7-8 20.7 a 7 20.7 a 7-9 20.6 a 21 20.2 b 17-18 19.3 c 6-8 19.1 c 4 12 3-12 2-12 5 10 20.9 2 0 . 8 2 0 . 8 20.6 a a a a Strength (10-* J) 2* 210 a 1 167 b Null 163 b 13-16 240 a 17-18 179 b 7-9 170 bc 7-8 165 bc 7 155 bcd 6-8 144 cd 21 127 d 5-10 3-12 2-12 4-12 197 160 152 150 a b b b Pelshenke (min) 2* 145 a 1 111 b Null 100 b 13-16 135 a 21 129 b 17-18 116 bc 7-9 110 bcd 7-8 104 bcd 6-8 90 cd 7 72 d 5-10 2 12 3-12 4-12 144 82 70 67 a b b b HMW-GS Effect (%) X.ij 2* 48.7 1 24.0 Null - 13-16 49.6 7-9 29.8 17-18 26.3 7-8 26.8 7 9.5 21 4.2 6-8 - 5-10 3-12 2-12 4-12 38.0 3.1 -3.0 61 p o r t i o n a l to the allelic effects are shown in Table 6. They allowed us to calculate the index IQ o f any genotype having 3 of the 16 alleles. F o r example, a genotype having the following H M W subunits o f glutenins 1 - 2 7 - 8 - 1 2 would have an index IQ = 40. The index IQ is similar to the gluten score (GS) proposed by Payne et al. (1987) from several sets o f offsprings and varieties assessed both by SDS-Page and by the SDS sedimentation test. Coefficients were established for the gluten score from the 14 alleles (i.e. 3 o f G l u - A l , 7 o f Glu-B1 and 4 o f Glu-D1), and 84 genotypes differing by their H M W allelic combination could be created. A highly significant correlation (r=0.881, n = 8 4 ) was found between IQ and GS. The index IQ calculated from the three parameters o f rheology (strength, tenacity and swelling) and the Pelshenke test was c o m p a r e d to both the six selection indices and to the technological test. Comparison o f the indices The indices were applied to F 4 offspring of 30 lines displaying a wide variation for their technological values (Table 7). The correlations calculated from the phenotypic values showed strong relationships between the four indices IW1, I W / , IV a and IV 2 . The index IQ was not so well related to the technological indices and was not significantly correlated to I W 3 . The correlations were even lower when GS was used (data not shown). The potentialities o f quality from the index IQ do not correspond to the entire variability o f the quality estimated by the indices I W i or IVi . C o m p l e m e n t a r y results arose when we analysed the number of c o m m o n lines selected both by direct selection using one p a r a m e t e r and by the technological indices: 1) when grain protein content was used alone for direct breeding, no more than 60% o f the lines selected by the indices were retained (Table 8); 2) all lines selected by the indices which included the Pelshenke test were also retained when this test was used alone; 3) the mixograph parameters should have selected no more than 80% o f the lines obtained from the indices; 4) the index IQ globally selected 50% of the total lines retained by the indices. The values of the technological parameters were normally distributed, except for the Pelshenke test, which revealed more low quality lines than the other tests. Con- Table 6. Value of coefficient Yij attributed to 16 alleles of the HMW-GS used in the quality index IQ Locus Alleles Glu-A 1 a 1* 15 b c 2* Null 30 0 Glu-B1 Glu-D1 " HMW-GS Quality coefficient a 7 b c d e f g h i j 7-8 7-9 6-8 20 13-16 13-19 14-15 17-18 21 18 20 2 2 32 18 5 8 a b c d e 2-t2 3-12 4-12 5-10 2-11 7 9 5 30 7 J Table 7. Statistical values of the technological parameters used to assess the quality of the 30 F4 lines and to calculate the indices and linear correlations between the seven indices (n = 30) Minima Maxima Mean (m) m + SE Pro Pel Zym MTM MHM MH7 MT7 MIN 12.95 16.85 14.54 14.71 18 200 69.9 78.2 25 53 39.1 40.3 1 4.3 2.3 2.4 6 9.4 7.7 7.9 5.5 7.8 6.6 6.7 1 1.9 1.3 1.3 19.3 44.0 29.2 30.2 IWl iw~ 1.ooo IW2 IW3 IV1 IV2 IV3 IQ 0.993"* 0.739 ** 0.972 ** 0.992"* 0.701 ** 0.571 ** IW2 IW3 IV1 1.000 0.720 ** 0.936 ** 0.998"* 0.728 ** 0.559"* 1.000 0.757 ** 0.698"* 0.618 ** 0.329 0.937"* 0.622"* 0.567"* *, ** Significantly correlated at 5% and 1%, respectively IV2 IV3 IQ 1.000 0.686"* 0.557"* 1.000 0.442" 1.000 1.000 62 GLU ( i w l ) IWl GLU (iv1) IVl i77.5 -851 ] 50 40 30 20 i I I0-20 [ , I 10 0 ---4 410 2r0 0 ~I0 80 I0050 40 GLU (iw2) IW2 i 30 20 10 0 0 20 40 ~60 8'0 100 GLU (iv2) IV2 65-75! I 45-551 35-45 i b t e i 1 t i 50403020 i 10 IW3 i 20 0 II 60 ] 1'0 0 () i 210 IV3 s 60 80100 GLU (iv3) ] 8~-87.5 t ] ] 82.8- 851 80- 82.51 ] ] 8__00! 177.5g +-- 801004 ~50 4 0 3 0 2 0 GLU(iw3) 187.5-901 [] 4~0 75- 7 7 . 5 ~ ! 1,7z5-7~1~ e 5~0 40 3'0 20 ~70-72.5 t ~ m67.5-70L - - 0 20 10 0 ] , ~0 , ---~] f 60 -- 80 10050 ~ 40 310 20 I0 0 0 20 40 60 80 100 Fig" 1 a-f. Distribution of the values of the six selection indices (leftside) observed from the 30 F 4 lines and the corresponding frequency of the lines having their quality index IQ higher than m + S E (rightside), a IWt, b IWz, e IW3, d IVI, e IV2, f IV 3 . * Limit value m + SE of the selection index Table 8. Percentages of the lines selected using the index (IWi or IVi) that were also independently selected with only one parameter. The m + SE limit values for selection indices and the parameters are given in Fig. I and Table 7, respectively Indices IW 1 IW e IW 3 IV 1 IVz IV3 Pro 41.6 41.6 60.0 45.5 41.6 33.3 Pel lO0.O 100.0 70.0 100.0 100.0 66.6 Zym 83.3 83.3 80.0 90.9 83.3 58.3 Mixo IQ MTM MHM MH7 MT7 MIN 83.3 83.3 60.0 81.8 83.3 75.0 58.3 58.3 80.0 63.6 58.3 58.3 50.0 50.0 80.0 54.5 50.0 66s 58.3 58.3 70.0 63.6 58.3 66.6 75.0 75.0 60.0 81.8 75.0 66.6 50.0 50.0 50.0 54.5 50.0 50.0 63 sequently, the distributions of IW 1 , IW2, IV 1 and IV2 indices which included the Pelshenke test, were asymmetrical (Fig. I a, b, d and e respectively). Indices IW3 and IV3 were normally distributed. The proportions of the lines having IQ > m + SE in each quality class of the indices IWi or IVi showed the following (Fig. 1). (1) All lines having the highest IWi or IVi values also had a high IQ. (2) On the other hand, many lines selected with IQ (IQ > m + S E ) would have been discarded through the selection indices (IW~ or IV~ < mi + SE~). The proportions of these lines expressed as percent of discarded lines is the same for all the selection indices and is about 54%. (3) The proportions of the lines selected by using the selection indices, which then should have been discarded by the index IQ (IQ < m + SE), were the following: IW 1 : 41.7%; IW2: 41.7%; IW3: 40%; IV1: 30%; IV2: 41.7% and IV3: 50% (expressed in percent of the selected lines). Although the index IQ was correlated with five indices, the potentialities of quality it evaluated do not correspond to all of the variability measured by the different parameters used in the selection indices. Discussion The quality indice IQ was strongly correlated with the gluten score GS published by Payne et al. 1987; Payne 1987. The GS was essentially established from comparisons of SDS sedimentation values, whereas IQ was calculated from different parameters of rheology and from dough swelling time of the Pelshenke test. The allelic effects, which correspond to the average of the differences between the quality values of the genotypes having a given allele and those having the reference allele, represent the true additive effects attributable to present-day European bread wheats. These effects would have been different if the genetic origin of the 195 wheats was less diversified. The simple statistical method used for calculating the IQ coefficient can be used for developing other IQs more adapted t o the cultivars and to the quality objective of a given country. Few cultivars had the alleles Glu-Ble (band 20) and Glu-DIe (bands 2-11). Consequently, these alleles were not introduced into the comparison of the means (Table 5). Their effects were estimated by comparing the quality values of the cultivars differing only by their Glu-Bl or Glu-D1 alleles. Only two cultivars had the rye secalins; consequently, the effect of the I B L / I R S translocation on dough stickiness and weakness (Dhaliwal et al. 1988) was not reported. Using the same parameters we have studied the effects of the 1BL/1 RS translocation on another set of cultivars and calculated the coefficient attributed to the indice IQ (unpublished). Many studies have been conducted to evaluate the efficiency of the index of selection in several crops for improving different characters: grain yield in maize (Robinson et al. 1951), wheat (Smith 1936), barley (England 1977) and oats (Eagles and Frey 1974), and for many other genetic traits as well (see Baker 1986 for review). In many cases it was demonstrated that introducing the character to be improved as a parameter of the index increased the relative efficiency of the index. In our case the strength W and loaf volume LV were not introduced in the index. All of the parameters used in the six indices contributed to a significant explanation of the character W or LV to improve. Moreover, for the three tests Pel, Zym and Mixo at least one of their parameters had a greater expected response to increase in the mean genotypic values of W or LV when combined in the index than it should have if used alone. As the alveograph and the baking test were not performed on the 30 lines, the superiority of the genetic values for the W or the LV of the selected lines was not confirmed. Nevertheless, for IW 1 , IW 2 and IV1 or IV2, 75 100% of the selected lines using these indices were also selected by the parameters Pel, Zym and Min. These percentages dropped to 58 to 60% when the mixograph was combined with Pro. As the index integrates several parameters, each corresponding to different technological properties, the observed complementation of IWi or IVi with the index IQ seems to be obvious. To determine the maximum efficiency for breeding, in using first the HMW-GS analysis, then the technological evaluations, more studies are needed. However, our results confirm what P. Kolster (personal communication) showed by selection with the HMW-GS profile and the SDS sedimentation test: breeding on the HMW-GS profile may lead to the elimination of many lines that would not have been discarded using a technological test. Consequently, for minimizing the losses of good quality lines when screening first from HMW-GS, it seems advisable to retain genotypes of IQ values higher than the mean m of the population rather than m + SE. Several criteria may be more important than others in choosing the most appropriate index from the six selection indices studied. From 10 to 25 g of grain are needed for the technological tests used in our indices. For the small-scale computerized mixograph, which helps wheat breeders test earlier generations (Rath et al. 1990), these amounts can be reduced. The mixograph initially developed for clarifying the dough characteristics in order to produce an optimum loaf (Swanson and Working 1933; Finney and Shogren 1972) also has parameters both correlated to strength and heritable (Branlard et al. 1991). However, the mixograph is less capable of evaluating the poor quality wheats than the Pelshenke test. Thus, there are some advantages to using indices IW2 and IV2. 64 Conclusion Each year breeders use several technological tests for assessing quality without relating them in a selection index that would provide more information on the genetic value of their material. F o r the first time breeding indices are p r o p o s e d for the quality evaluation of bread wheats. According to quality objectives, these indices can be used in early generations in addition to H M W - G S determination. The indices I W and IV m a y be combined in a global index I = p I W + qIV where p = ] - q is chosen according to the priority of breeding. M a n y other selection indices could be created for wheat quality; among these, the introduction of the SDS sedimentation test as an alternative to the Zeleny, and grain hardness could be very useful for the further improvements. Acknowledgements. French private breeders belonging to Club 5, for their active participation to that study, and Dr. T. S. Cox, for reading the manuscript, are gratefully acknowledged. References Afnor Norme exp6rimentale V-03-713. Association Franqaise de Normalisation, Tour Europe Cedex 7, 92049 Paris Baker RJ (1986) Selection indices in plant breeding. CRC Press, Boca Raton, Fla. Baker RJ, Tipples KH, Campbell WP (1971) Heritabilities of food and correlations among quality traits in wheat. Can J Plant Sci 51:441-448 Bietz JA, Shepherd KW, Wall JS (1975) Single kernel analysis of glutenin: use in wheat genetics and breeding. Cereal Chem 52:513 532 Branlard G, Dardevet M (1985 a) Diversity of grain proteins and bread wheat quality. I. Correlation between gliadin bands and flour quality characteristics. J Cereal Sci 3:329-343 Branlard G, Dardevet M (1985 b) Diversity of grain proteins and bread wheat quality. II. Correlation between high molecular weight subunits of glutenin and flour quality characteristics. J Cereal Sci 3:345 354 Branlard G, Rousset M, Loisel W, Autran JC (1991) Comparison of 46 technological parameters used in bread wheat quality evaluation. J Genet Breed 45:263-280 Cox TS, Shogren MD, Sears RG, Martin TJ, Bolte LC (1989) Genetic improvement in milling and baking quality of hard red winter wheat cultivars, 1919 to 1988. Crop Sci 29:626631 Dhaliwal AS, Mares DJ, Marshall DR, Skerritt JH (1988) Protein composition and pentosan content in relation to dough stickiness of 1B/1R translocation wheats. Cereal Chem 65: 143-149 Eagles HA, Frey KJ (1974) Expected and actual gains in economic value of oat lines from five selection methods. Crop Sci 14:861-864 England F (1977) Response to family selection based on replicated trials. J Agric Sci 88:127-134 Finney KF, Shogren MD (1972) A ten gram mixograph for determining and predicting functional properties of wheat flour. Baker's Dig 46:32-42 Fowler DB, de la Roche IA (1975) Wheat quality evaluation 2. Relationships among predicting tests. Can J Plant Sci 55: 251-262 Motto M, Perenzin M (1982) Index selection for grain yield and protein improvement in an opaque - 2 synthetic maize population. Z Pflanzenziicht 89:47-54 Payne PI (1987) The genetical basis of breadmaking quality in wheat. Aspects Appl Biol 15:79 90 Payne PI, Lawrence GJ (1983) Catalogue of alleles for the complex gene loci Glu-A1, Glu-B1 and Glu-D1 which code for high-molecular-weight subunits of glutenin in hexaploid wheat. Cereal Res Commun 11:29-35 Payne PI, Corfield KG, Holt LM, Blackman JA (1981) Correlation between the inheritance of certain high-molecularweight subunits of glutenin and breadmaking quality in progenies of six crosses of bread. J Sci Food Agric 32:51-60 Payne PI, Nightingale MA, Krattiger AF, Holt LM (1987) The relationships between HMW glutenin subunit composition and the bread-making quality of British-grown wheat varieties. J Sci Food Agric 40:51-65 Rath CR, Gras PW, Wrigley CW, Walker CE (1990) Evaluation of dough properties from 2 g flour using the Mixograph principle. Cereal Foods World 35:572-574 Robinson HF, Comstock RE, Harvey PH (1951) Genotypic and phenotypic correlations in corn and their implications in selection. Agron J 43:282-287 Smith HR (1936) A discriminant function for plant selection. Ann Eugen 7:240-250 Sozinov AA, Poperelya FA (1980) Genetic classification of prolamines and its use for plant breeding. Ann Technol Agric 29:229 -245 Swanson CO, Working EB (1933) Testing the quality of flour by the recording dough mixer. Cereal Chem 10:1 29