Genetic parameters of production traits in Atlantic salmon (Salmo salar)

Genetic parameters of production traits in Atlantic salmon (Salmo salar )Joseph Powell a,⁎,Ian White b ,Derrick Guy c ,Sue Brotherstone baRoslin Institute,Roslin,Midlothian,Scotland,EH39PS,UKbUniversity of Edinburgh,The School of Biological Sciences,Ashworth Labs,King's Buildings,Mayfield Road,Edinburgh,EH93JR,UKcLandcatch Natural Selection Ltd,The e-Centre,Cooperage Way,Alloa,Clackmannanshire,FK103LP ,UKReceived 13April 2007;received in revised form 16November 2007;accepted 23November 2007AbstractPhenotypic and genetic parameters of biometrical and carcass traits from two pedigreed populations of an Atlantic salmon breeding program were estimated using restricted maximum likelihood.Each of the populations (broodstock and commercial sib test)comprised of the same 200full-sib families.Heritability estimates for biometrical traits (sw1wt,sw2wt,harvwt,filletwt,harvlen,guttedwt,deheadwt,guts,head,and carcass)were low to high (0.12–0.53);estimates for fillet fat %and colour traits (hue,saturation,and intensity)were low to moderate (0.16–0.33).The heritabilities for yield traits (guttedwt%,deheadwt%,filletdeheadwt%,and fillet%)measured as a ratio of weights,were low (0.009–0.037),due to the two traits being a fixed proportion of each other.In the broodstock population significant sex (P b 0.001)and tank (P b 0.01)effects were found for first and second sea winter weights (sw1wt and sw2wt)with male fish on average 0.14kg heavier at sw1wt and 2.34kg heavier at sw2wt.Estimated genetic correlations (r G )between biometrical traits and fillet fat percentage were all positive and ranged from 0.34to 0.82.Selection for increased harvest weight is expected to produce favourable changes in fillet weight (r G =0.99)but unfavourable changes in fillet fat percentage (r G =0.80).Favourable estimates of the genetic correlation between saturation (colour score)and biometrical traits were all positive (r G =0.23–0.61).©2008Elsevier B.V .All rights reserved.Keywords:Heritability;Genetic correlations;Atlantic salmon;Salmo salar1.Introduction1.1.Atlantic salmon in aquacultureAtlantic salmon (Salmo salar )is one of the most established and economically important species within aquaculture and,as a result,a number of studies on the genetic and phenotypic parameters of productions traits have already been conducted (Gjerde and Gjedrem,1984;Standal and Gjerde,1987;Gjerde et al.,1994;Norris and Cunningham,2004;Rye and Refstie,1995;Rye and Gjerde,1996;Quinton et al.,2005).Despite this,there is little consistency in the traits analysed or results obtained,leading to the suggestion that genetic parameters are highly specific to a study population (Kause et al.,2002).Important traits of aquaculture species are manipulated through breeding schemes that are employed to maximise the genetic gain in addition to controls over diet and culture environment,harvesting procedures and timing of the product delivery to the market (Paterson et al.,1997).Understanding the interactions between the economically important traits is of particular importance if we are to control the product quality in terms of body composition and flesh quality.As farming of Atlantic salmon is growing as an aquaculture industry,the need to estimate the genetic (co)variances of traits is of increasing importance.Quantitative genetic studies are required to design effective breeding programmes,specific to a given population,which are aimed at improving the economic value and quality of the fish.Previous genetic studies have revealed moderate levels of genetic variation for production traits —such as body and fillet weight and flesh colour —of salmonids that allow genetic improvement through implementation of selection programmes (Friars et al.,1995;Gjedrem,2000;Hulata,2001).Available online at Aquaculture 274(2008)225–231/locate/aqua-online⁎Corresponding author.Tel.:+441315274200;fax:+441314400434.E-mail address:joseph.powell@ (J.Powell).0044-8486/$-see front matter ©2008Elsevier B.V .All rights reserved.doi:10.1016/j.aquaculture.2007.11.0361.2.Important traits within aquacultureBody weight and growth rate have generally been the traits assigned the highest importance in terrestrial livestock selection programmes and this is also the case for Atlantic salmon (Quinton et al.,2005).The harvest weight at2–3years for Atlantic salmon has generally been found to have moderate heritabilities(Gjerde and Gjedrem,1984;Standal and Gjerde, 1987;Gjerde et al.,1994;Rye and Refstie,1995;Quinton et al., 2005)and has therefore been a trait of considerable importance in selection rger fish,at time of harvest,are more desirable for a salmon farmer;however,the effect of selecting for increased body weight on other production traits is not fully understood and needs to be considered in designing breeding programmes to predict how changes in growth rate affect other commercially important traits(Gjedrem,2000).The quality of the product at the time of harvest affects the economic value of the fish and is generally determined by the body composition and carcass quality traits,such as meat colour and fat content.The importance of product quality means that most breeding programs for salmonid species should incorporate quality traits along with body weight as selection criteria (Gjedrem,1997,2000).The market considers flesh colour an indicator of salmon quality and retailers will downgrade or even reject a product with insufficient colour.The distinctive pink-red colouration of the flesh is caused by carotenoid pigments that salmonids are unable to synthesise.Wild salmon obtain these from a diet of amphipods and crustaceans;however,such a diet is unsuitable for farmed salmon and so pigments are usually added to their feed in the form of astaxanthin and canthaxanthin.These pigments are an expensive component of salmonid feed so,to minimise costs,producers want animals that most efficiently absorb and retain these pigments in the flesh.Flesh pigmentation has thus become a trait of interest for breeding programmes.A number of studies in salmonids have reported positive genetic correlations between body weight and colouration scores, indicating that selection for increased weight will result in an increase in the desirable colouration of the flesh(Norris and Cunningham,2004;Quinton et al.,2005;Johnston et al.,2006). Another quality trait of increasing interest is fillet fat content.An excessive amount of fat in muscles is thought to have a detrimental effect on meat texture as well as affecting smoking processes and quality traits such as colouration.As a general rule fillet fat percentage greater than18%is considered undesirable(Gjedrem, 1997).High fat content is also thought to affect the processing of the carcass,making handling more difficult and errors more likely to occur(Rye and Gjerde,1996).There is evidence that direct selection for harvest body weight results in unfavourable indirect responses in fat content,as heavier fish have a greater fillet fat percentage(Gjerde and Gjedrem,1984;Rye and Gjerde,1996; Quinton et al.,2005).Positive genetic correlations between fat content and body weight imply that if breeding programmes are going to maintain the quality of their product,they will need to utilise methods,such as restricted selection indices(Cameron, 1997)in order to control the expected increase in fat percentage.The objective of this study was to obtain estimates of genetic and phenotypic parameters for production traits in Atlantic salmon.2.Materials and methods2.1.Fish populationData were available from a commercial breeding programme where fish are maintained in a combination of freshwater and seawater tanks to mimic the life cycle stages of Atlantic salmon.The data represent1year class of a commercial strain of Atlantic salmon originating from a small number of founder strains available in Northern Europe.Data were collected from two Atlantic salmon populations,termed broodstock and sentinel.The broodstock population was composed of200full-sib family groups generated by mating each male with a maximum of4–5females and each female to a single male.Broodstock were spawned between October2001and December2001and initially hatched over a period of2weeks in March2002,then reared in individual family freshwater tanks.At about6months after hatching an electronic Passive Integrated Transponder(PIT)tag,which contains an individual identification number,was inserted into the abdominal cavity of each fish,approximately80per broodstock family and25per sentinel family.After tagging,broodstock families were mixed and reared in three freshwater tanks over winter until transfer to six seawater tanks in April2003.Measurements of body weight and other characteristics were taken at key stages as fish were moved between tanks.The salmon maturing after two winters in seawater were used as parents for the subsequent generation.The sentinel population were sibs taken at random from the main broodstock population,composed of individuals from the same200full-sib families and raised under normal commercial conditions.Sentinel fish were initially raised as separate families in the same freshwater tanks as the broodstock for6months and were then moved as mixed families to two communal freshwater tanks at the point their weight was taken at pit tagging.In April2003at approximately500days of age they were transferred to a single sea cage off the west coast of Scotland.There they remained with minimal handling until April2004,when they were harvested at approximately23–24months of age at an average weight of2kg.At all times during the rearing process all tanks and cages received identical treatment in terms of feed,vaccinations,and temperature and light regimes.The pedigree was constructed from a combination of family tank records and microsatellite markers.Genetic groups were assigned based on the founder strains of fish.These were used to differentiate between animals in the pedigree with missing parentage in different years and preventing them from all being put into the same group.2.2.Measurement of traitsInitially the total number of broodstock and sentinel fish with pit tags was 18,135and6338respectively.Over the course of the breeding programme, random mortality and deliberate non-random culling(e.g.to skew sex ratios for mating)resulted in the number of recorded measurements for traits decreasing as the breeding programme progressed.(Taking account of non-random culling is partly the motivation for multivariate estimations).Although mortalities from the sentinel population were minimal,as the harvested population was processed the number of measurements decreased at random through natural losses and instrumental errors.2.2.1.Traits recorded on the broodstock populationTwo traits were measured in the broodstock population;one sea winter weight(sw1wt),taken in February2004when the fish were approximately 24months old,and two sea winter weight(sw2wt),taken in February2005 when the fish were approximately36months old.Weights were measured to the nearest10g.The age(in days from first feeding)of each fish at weighing was also recorded.The specific tank and sea cage that broodstock were reared in as recorded are included as fixed effects in the models for sw1wt and sw2wt.The sex of broodstock fish was recorded as was determined at any stage during the breeding programme.The sex of many fish remained indeterminate,due to the relatively late stage at which sexual development occurs in salmonids,and these fish have their sex recorded as unknown.2.2.2.Traits recorded on the sentinel populationAll traits on the sentinel population were recorded at the time of harvesting. Data on harvest performance of fish were obtained from a commercial processing operation using automated equipment.Fish were first weighted(harvwt)and their226J.Powell et al./Aquaculture274(2008)225–231length measured from the apex of the tail fin to the nose (harvlen).The fish were then processed,and at each stage of processing weights were taken of the removed component and the remainder of the fish.Fish were first gutted (guttedwt and guts)and then deheaded (deheadwt and head)before removing and weighing both sides of the fillet (filletwt)leaving the remainder of the fish as waste (waste).All weights were measured to the nearest 10g and lengths to the nearest cm.Four yield traits were calculated from the measured weights;gutted weight as a percent of harvest weight (guttedwt%),deheaded weight as a percent of gutted weight (deheadwt%),fillet weight as a percent of deheaded weight (filletdehead%)and fillet weight as a percentage of harvest weight (fillet%)(Table 1).The fat percentage (fatpc)of the gutted fish was calculated as the integrated mean of 8readings taken at various positions along the body,using the Torry Fatmeter (Distell Ltd).The colour parameters (hue,saturation and intensity)were extracted from digital images taken of the fillet,with a FinePix S602camera,using ‘ImagePro Express ’software (Media Cybernetics Inc).The ‘HSI ’colour model was used,where hue is a polar co-ordinate measure of ‘redness ’on the scale of 0–90°(red to yellow)with values closest to zero being desirable.Saturation,equivalently called chroma,is the colour saturation as a deviation from white to fully saturated (0to 225),with low score being washed out and high scores the preferred deeper colour.Intensity is approximately the opposite of saturation as a deviation from fully saturated to black (0to 225).As opposed to saturation,low scores are preferable although there is little information in intensity not inherent in saturation,and the HIS essentially resolves to a two-parameter model.Gutting was by machine which precluded observation of sexual develop-ment in these pre-breeding stage fish,and so the sex of the fish was not included in any of the models for traits measured in this population.The sentinel population was reared in a single cage,and therefore the cage was not included within the models for any of the traits measured on this population.Age (in days from first feeding)of each fish was recorded when each measurement was taken.2.3.Data analysisThe analysis of the data is divided broadly into two sections:a univariate analysis in which heritabilities for yield traits (guttedwt%,dehead%,filletde-head%and fillet%)and additional biometric traits (gut,head and waste)were estimated;and a multivariate analysis in which genetic and phenotypic (co)variances were estimated between biometric (sw1wt,sw2wt,harvwt,filletwt,harvlen,guttedwt and deheadedwt)and carcass quality traits (fatpc,hue,saturation and intensity).The software package ASReml (Gilmour et al.,1998)was used for all analyses.2.3.1.Univariate analysisGenetic and phenotypic parameters of the yield traits (guttedwt%,dehead%,filletdehead%,and fillet%),gut weight (gut),head weight (head),and waste weight (waste)were estimated with a univariate animal model (Eq.(1)).These traits were initially included in the multivariate model but caused convergence problems,due to high correlations between traits,and so were removed and analysed individually.Age was included as a linear covariate for each trait but was found not significant for all traits,reflecting the narrow hatching window of 2weeks and subsequently dropped from all further analyses.As sex was not determined and there was only a single tank effect,these were not fitted as effects in the models for sentinel data.y j ¼A þb age j þa j þe jð1Þwhere y j is the observation for individual j ,μis the population mean,βis the regression coefficient on the age (age)of individual j ,a j is the genetic effect for animal j ,and e j is the random residual error for individual j .Common environment effects potentially exist in the data.The most important relates to the individual ‘family unit ’tank environment for a short period following hatching,confounded with full-sib family.There was no replication of families over tanks at this stage and so the effect was not possible to estimate from this data.However,after tagging,families were mixed and replicated over the available tanks allowing the tank effect to be accounted for explicitly.We made no attempt to include a maternal environmental component in the model as there was little information from which to estimate it and it is expected to be unimportant in a species where maternal and paternal involvement ceases once eggs are stripped.A pedigree file tracing back to the founder strains was included in the analyses.2.3.2.Multivariate analysisA multivariate analysis was performed for biometric traits (sw1wt,sw2wt,harvwt,filletwt,harvlen,guttedwt,deheadedwt and waste)and carcass quality traits (fatpc,hue,saturation and intensity).The fixed effects of sex and tank/sea cage were only measured in the broodstock population,and so in the multivariate model these fixed effects will only apply to sw1wt and sw2wt.The random effects of the animal and age as a linear covariate for each trait wereTable 1Number of observations (n ),mean and standard deviation (S.D.)for the traits analysed,plus a description of each and on which population they were measured Trait Trait type Descriptionn Mean S.D.Broodstock Sw1wt Biometric One sea winter weight (kg)15,724 2.670.87Sw2wt BiometricTwo sea winter weight (kg)835210.12.46Sentinels Harvwt Biometric Weight at harvest (kg)3172 1.980.75Filletwt Biometric Weight of fillet (kg)1761 1.410.54Harvlen Biometric Length at harvest (cm)317255.6 6.77Guttedwt Biometric Weight of gutted fish (kg)1555 1.810.67Deheadwt Biometric Weight of deheaded fish (kg)1829 1.610.59Guts Biometric Weight of guts (kg)15150.260.17Head Biometric Weight of head (kg)14600.210.11WasteBiometric Weight of residual waste (kg)17170.200.09Guttedwt%Yield Gutted weight/harvest weight (%)155489.00.13Deheadwt%Yield Deheaded weight/gutted weight (%)150890.00.13Filletdehead%Yield Fillet weight/deheaded weight (%)172787.00.04Fillet%YieldFillet weight/harvest weight (%)176069.00.08Fatpc Carcass quality Fat percentage of whole fish (%)168213.5 4.26HueCarcass quality Redness measure (scale 0–90)177013.30.75Saturation Carcass quality Colour saturation as deviation from white (scale 0–225)177013917.9IntensityCarcass quality Colour saturation as deviation from black (scale 0–255)1770116 6.74227J.Powell et al./Aquaculture 274(2008)225–231included.The following multivariate animal model was used to estimate the(co) variances of traitsy ij¼A iþs ikþt ilþb i age jþa ijþe ijð2Þwhere i represents the traits sw1wt,sw2wt,harvwt,filletwt,harvlen,guttedwt, deheadedwt,fatpc,hue,saturation and intensity,μi is the population mean for trait i,y ij is the observation of trait i for animal j,s ik is the fixed effect of sex k on trait i,t il is the fixed effect of tank/sea cage l on trait i andβi is the linear coefficient of age(age)on trait i.a ij is the random genetic animal effect on trait i for animal j,and e ij is the random residual error of trait i for animal j.Convergence problems caused by high genetic correlations between traits meant that a single multivariate analysis was not possible.Multivariate analyses on subsets of traits were therefore performed and the results assembled into a single correlation matrix(Table2).ASReml was used to estimate heritabilities, as well as phenotypic(r p)and genetic(r G)correlations,along with their standard errors.3.Results3.1.Descriptive statisticsThe total number of measurements recorded for each trait and their phenotypic means and standard deviations are given in Table1.Sex and tank were fitted as fixed effects for the first and second sea winter weights of the broodstock population(sw1wt and sw2wt).Very highly significant(P b0.001)sex effects,with male fish on average 0.14kg heavier at sw1wt and2.34kg heavier at sw2wt,and highly significant tank effects(P b0.01)were observed for both first and second sea winter weights.No significant age effects were detected for any of the traits.The results are divided into two sections;univariate analysis of yields and processing traits,and a multivariate analysis of the biometric and carcass quality traits.3.2.Univariate analysis of yield and processing traitsVariance components and heritabilities for yields and processing weights are given in Table3.Heritabilities for all yield traits were very low(0.009–0.037)and none are significantly different from zero.The heritabilities of head and guts were low and of waste moderate.Table2Genetic correlations on the lower triangle and phenotypic correlations on the upper triangleSw1wt Sw2wt Harvwt Filletwt Harvlen Guttedwt Deheadwt Fatpc Hue Saturation Intensity Sw1wt0.510.740.530.590.440.580.610.320.040.150.07(0.04)(0.008)(0.036)(0.036)(0.040)(0.041)(0.039)(0.040)(0.041)(0.045)(0.043) Sw2wt0.840.430.320.340.270.370.330.130.010.090.05(0.024)(0.04)(0.038)(0.038)(0.038)(0.042)(0.042)(0.040)(0.036)(0.042)(0.039) Harvwt0.760.580.500.970.85⁎0.980.05−0.330.44−0.04(0.043)(0.063)(0.05)(0.001)(0.006)(0.001)(0.019)(0.023)(0.022)(0.027) Filletwt0.760.580.990.520.88⁎⁎0.48−0.330.49−0.06(0.045)(0.063)(0.0014)(0.055)(0.006)(0.026)(0.023)(0.022)(0.027) Harvlen0.850.580.950.960.460.890.890.39−0.350.45−0.004(0.034)(0.068)(0.0124)(0.011)(0.05)(0.006)(0.005)(0.022)(0.024)(0.023)(0.028) Guttedwt0.810.67⁎⁎0.970.510.980.48−0.350.43−0.04(0.0466)(0.064)(0.01)(0.06)(0.001)(0.027)(0.024)(0.024)(0.029) Deheadwt0.810.550.99⁎0.970.990.530.46−0.330.45−0.06(0.038)(0.071)(0.001)(0.0087)(0.004)(0.055)(0.022)(0.023)(0.022)(0.027) Fatpc0.760.340.800.820.760.780.790.28−0.220.31−0.05(0.067)(0.107)(0.056)(0.062)(0.066)(0.072)(0.06)(0.05)(0.031)(0.031)(0.033) Hue−0.41−0.19−0.47−0.46−0.52−0.50−0.5−0.300.160.02−0.01(0.11)(0.135)(0.116)(0.116)(0.112)(0.113)(0.117)(0.167)(0.043)(0.026)(0.026) Saturation0.340.230.5120.610.510.480.540.58−0.490.33−0.64(0.09)(0.108)(0.085)(0.077)(0.07)(0.101)(0.088)(0.11)(0.148)(0.056)(0.015) Intensity0.130.14−0.10−0.15−0.07−0.05−0.19−0.080.43−0.830.26(0.112)(0.116)(0.116)(0.12)(0.123)(0.126)(0.124)(0.158)(0.16)(0.053)(0.052) Heritabilities are shown in bold on the diagonal.⁎denotes the correlations that we were unable to estimate due to convergence problems.Table3Genetic variance(V G),residual variance(V E)and heritability(h2),plus the standard errors of the heritabilities(S.E.)for the yield and processing traitsTrait V G V E h2(S.E.)Guttedwt%0.036 1.5900.02(0.031) Dehead%0.014 1.6300.01(0.025) Filletdehead%0.0070.1820.04(0.026) Fillet%0.0180.6710.03(0.023) Gut(weight)0.392 2.8420.12(0.040) Head(weight)0.136 1.0250.12(0.039) Waste(weight)0.695 1.6060.30(0.057)The variance components have been multiplied by100.Fig.1.The natural log of fillet weight is plotted against the natural log of deheaded weight.The ratio of these traits on the original scale is filletdehead%.228J.Powell et al./Aquaculture274(2008)225–231These low heritabilities for yield traits are undesirable in terms of breeding objectives and this may be due to the way the traits were calculated.If the two traits used to calculate the yields are in approximate proportion to one another then taking a ratio of the traits will result in very little genetic variation,and hence very low heritabilities.To test this,natural logs of the weight traits,used to calculate yield ratios,were regressed against one another and an estimate of the slope of the line and intercept obtained.The data was log transformed before the analysis so that if one of the weight traits is a fixed proportion of the other then the regression of the log of one against the log of the other will give a regression line with a slope equal to1as was found in this dataset.An example of the graph of the regression for filletdehead%is given(Fig.1).Linear relationships between the natural logs of the two traits that comprise each of the yield traits were found.This indicates that one of the traits is approximately a fixed proportion of the other.Regression coefficients for the log log plots of the component traits for each of the yields are shown in Table4.3.3.Multivariate analysis of biometric and carcass quality traitsGenetic and phenotypic correlations,along with heritabilities are given in Table2.The heritabilities of the biometrical traits are all moderate to high,while the carcass quality traits have low to moderate heritabilities.There are large positive genetic correlations between the biome-trical traits,with some pairs of traits having very high correlations, notably fillet and harvest weight,harvest length and harvest weight, and harvest length and fillet weight.Fillet fat percentage has high positive correlations with most of the biometrical traits.Moderate to high genetic correlations were estimated between saturation and biometrical traits and with fat percentage.Favourable negative correlations between hue and biometrical traits and between hue and fat percentage were also estimated.Low genetic correlations are estimated for intensity with most traits,aside from hue and saturation which have higher correlations.4.Discussion4.1.Heritabilities4.1.1.Yield and processing traitsHeritabilities for the weights of components of the carcass (guts,head,and waste)measured in a univariate analysis are low to moderate(Table3).Comparison of the heritabilities reported here is difficult as we can find no other published reports of genetic parameters for the same carcass processing traits as estimated here,for Atlantic salmon.Within livestock breeding programmes the yield of the final product is a highly desirable trait,and is commonly a target for selection.In salmon the fillet is the only economically valuable part of the fish and processors wish to maximise the weight of the fillet for any given total weight of fish since that represents the profit margin for that sector of the industry.The definition of yield differs between studies and it appears that the methods used to calculate the yields in this study have resulted in very low heritability estimates.These low estimates are caused by the traits used to calculate the yields being an almost constant proportion of each another.Rutten et al.(2004,2005)also question the use of yield traits in selection criteria,after estimating very low genetic correlations between fillet yield and body weight,and low variance in yield traits.Alternative methods of estimating fillet yield need to be developed to get around the problem of high genetic correlations between weights of different components of the carcass.Such alternative methods might result in improved heritabilities for yield.Alternative approaches using for example ultrasound imaging or body composition analysis using scanning technology as used in breeding of terrestrial species may be appropriate in salmon.Within aquaculture species there are few reported estimates of the genetic components of fillet percentage or yield traits that are comparable to those reported here.However,Kause et al. (2002)estimated heritability for fillet weight,independent of genetic variation in body weight,of0.03for rainbow trout. They suggest that this reveals no genetic variation in fillet weight independent of genetic variation in body weight.Neira et al.(2004)estimated heritability for fillet percentage of0.11in Coho salmon,which is lower than their estimate of0.24for body weight and0.18for total fillet weight.It is likely that the situation may be different in other species making comparison with estimates given here difficult.However,there is no suggestion from this data that an increase in live weight will necessarily result in a greater proportionate increase in the non-edible parts of the carcass.4.1.2.Biometric and carcass quality traitsEstimates of heritabilities for the biometric traits(sw1wt, sw2wt,harvwt,filletwt,harvlen,guttedwt,and deheadedwt) measured in a multivariate analysis were moderate to high and heritabilities for the carcass quality(fatpc,hue,saturation,and intensity)traits were low to moderate(Table2).Although the estimates of heritabilities for biometrical traits reported here are higher than those published from other studies on Atlantic salmon (Gjerde and Gjedrem,1984;Standal and Gjerde,1987;Gjerde et al.,1994;Quinton et al.,2005)it is worth noting that there is considerable variation in these published estimates.This variability in heritability estimates(and genetic correlations) may result from genetic differences between the study popula-tions(Roff,1997;Roff and Mousseau,1999)and from differences in the environmental conditions in which different populations are raised(Hoffman and Merila,1999).Genetic differences between populations may be due to the high fidelity with which salmon home to their natal streams to spawn,which has resulted in significant levels of genetic structuring over local spatial scales(Nilsson et al.,2001).As most farmed salmon populations are derived from a relatively restricted founding stock,and have been subjected to selection pressure for desirableTable4Regression coefficients of the log log plots for component traits of each of the yield traits along with their standard errorsTrait Plot RegressioncoefficientS.E.Guttedwt%Log of guttedwt vs log harvestwt0.96210.015 Dehead%Log of deheadwt vs log guttedwt 1.02850.009 Filletdehead%Log of filletwt vs log deheadedwt0.96820.007 Fillet%Log of filletwt vs log harvwt 1.01130.016229J.Powell et al./Aquaculture274(2008)225–231。