scholarly journals Categorization of birth weight phenotypes for inclusion in genetic evaluations using a Deep Neural Network

2021 ◽  
Author(s):  
Andre Ribeiro ◽  
Bruce L Golden ◽  
Matthew L Spangler
Keyword(s):  
Birth Weight ◽  
Data Collection ◽  
Real Data ◽  
Phenotypic Variance ◽  
Sigmoid Function ◽  
Average Correlation ◽  
Calving Ease ◽  
Potential Impact ◽  
Genetic Evaluations ◽  
Weight Data

Abstract Birth weight serves as a valuable indicator of the economically relevant trait of calving ease and erroneous data collection for birth weight could impact genetic evaluations for calving ease. The objective of the current study was to evaluate the use of deep neural networks for classifying contemporary groups based on the method used to generate birth weight phenotypes. Contemporary groups (CG; n = 120,000,000) ranging between 10 and 250 animals were simulated assuming 12 data collection and CG formation scenarios that could impact CG phenotypic variance including: weights recorded with a digital scale (REAL), hoof tape (TAPE), and those that were fabricated (FAB). The performance of 6 activation functions (AF; ReLu, sigmoid, exponential, ReLu6, Softmax, Softplus, Leaky ReLu, and TangH) were evaluated. Four hidden layers were used with 7 different scenarios relative to the number of neurons. Simulations were replicated 10 times. In general, accuracy (proportion of correct predictions) across AF and numbers of neurons were similar, with mean correlations ranging between 0.91 and 0.99. The AF ReLu, Sigmoid, Exponential and ReLu6 had the greatest consistency (mean pair-wise correlation among replicates) with an average correlation of greater than 0.85. Independent of the number of neurons used, the sigmoid function produced the highest accuracy (0.99) and consistency (0.93). The model with the greatest accuracy and consistency was then applied to real birth weight data supplied by the American Hereford Association. In the real data, the lowest phenotypic variance was for FAB CG (2.65 kg 2), REAL CG had the largest (15.84 kg 2) and TAPE CG was intermediate (6.84 kg 2). To investigate the potential impact of FAB data on routine genetic evaluations, CG classified as FAB in 90% or more of the replicates were removed from the evaluation for calving ease and the rank of resulting genetic predictions were compared to the case were records were not removed. The removal of FAB CG had a moderate impact on the prediction of calving ease expected progeny differences, primarily for animals with intermediate to high accuracy. Results suggest that a well-trained DNN can be effectively used to classify data based on quality metrics prior to inclusion in routine genetic evaluation.

scholarly journals 21 President Oral Presentation Pick: Using Deep Neural Networks to determine birth weight data quality for genetic evaluations in beef cattle

2020 ◽  
Vol 98 (Supplement_4) ◽  
pp. 7-7
Author(s):  
Andre Ribeiro ◽  
Bruce L Golden ◽  
Matthew L Spangler
Keyword(s):  
Neural Networks ◽  
Birth Weight ◽  
Deep Neural Networks ◽  
Real Data ◽  
Phenotypic Variance ◽  
Training Procedure ◽  
Sigmoid Function ◽  
Average Correlation ◽  
Genetic Evaluations ◽  
Weight Data

Abstract The objective of this work was to evaluate the use of deep neural networks (DNN) for classifying contemporary groups based on the method used to generate birth weight (BWT) phenotypes. Contemporary groups (CG; n = 120,000) ranging between 10 and 500 animals were simulated assuming 12 data collection and CG formation scenarios that could impact CG phenotypic variance, including weights recorded with a digital scale (REAL), hoof tape (TAPE), and those that were fabricated (FAB). The performance of 6 activation functions (AF; ReLu, sigmoid, exponential, ReLu6, Softmax, Softplus) were evaluated. Four hidden layers were used with 7 different scenarios relative to the number of neurons. The training procedure was implemented in Python 3 with TensorFlow 1.14 and the ADAM optimization. Simulated CG were divided into training (80%) and testing (20%). The correlation between the observed and predicted CG types, averaged across 10 replicates, was used to assess accuracy and the correlations of predictions between replicates were used to measure the consistency of the model. In general, accuracy across AF and numbers of neurons were similar, with mean correlations ranging between 0.91 and 0.99. The AF ReLu, Sigmoid, Exponential and ReLu6 had the greatest consistency between the replicates, with an average correlation greater than 0.90. Independent of the number of neurons used, the sigmoid function produced the highest accuracy (0.99) and consistency (0.96). The DNN was retrained using 10-fold the number of CG and mimicking the CG size distribution observed in real data obtained from the American Hereford Association (n = 46,177 CG). In the real data, the lowest phenotypic variance was for FAB CG (2.98 kg2), REAL CG had the largest (18.33 kg2) and TAPE CG was intermediate (8.64 kg2). Results suggest that a well-trained DNN can be effectively used to classify data based on quality metrics prior to inclusion in routine genetic evaluation.

scholarly journals Evaluación fenotípica y genética para características de crecimiento en la raza criolla colombiana Costeño con Cuernos

2007 ◽  
Vol 7 (2) ◽  
pp. 12 ◽  
Author(s):  
Rodrigo Alfredo Martínez ◽  
Juan Esteban Pérez ◽  
Teófilo Herazo
Keyword(s):  
Birth Weight ◽  
Variance Components ◽  
Fixed Effects ◽  
Genetic Parameters ◽  
Growth Traits ◽  
Phenotypic Correlation ◽  
Phenotypic Variance ◽  
Direct Effects ◽  
Weaning Weight ◽  
Number Of Births

<p>Se establecieron componentes de varianza, así como parámetros fenotípicos y genéticos, respecto de las variables ‘peso al nacimiento’, ‘peso al destete’ (ajustado a los 270 días) y ‘peso a los 480 días’ en un hato del ganado criollo colombiano Costeño con Cuernos. Se analizaron 2.281 registros de pesos al nacer, 1.722 de pesos al destete  y 1.086 de pesos ajustados a los 480 días utilizando la metodología de máxima verosimilitud restringida (DFREML). También se ajustó un modelo animal que incluyó efectos genéticos directos, maternos y de ambiente permanente, asumiendo como efectos fijos el año de nacimiento, el sexo del ternero y el número de partos de la madre; finalmente, se estimaron los parámetros genéticos ‘heredabilidad’, ‘repetibilildad’ y se establecieron correlaciones genéticas y fenotípicas. Se reportan bajas estimaciones de heredabilidad de los efectos directos, que varían entre 0,17 ± 0,001 y 0,21 ± 0,074 para los pesos al nacer y al destete, respectivamente; así mismo, fue baja la heredabilidad de los efectos genéticos maternos con relación al peso al nacimiento, aunque estos estimados aumentaron respecto de los pesos al nacer y al destete. Las correlaciones entre efectos directos y maternos fueron negativas, pero el mayor valor se encontró para el peso al nacimiento (-0,89). La contribución del ambiente permanente como proporción de la varianza fenotípica total fue baja y disminuyó a medida que aumentó la edad del animal.</p><p> </p><p><strong>Genetic and phenotypic evaluation to characterize growth traits of the native Colombian breed Costeño con Cuernos</strong></p><p>For a herd of native Colombian breed of cattle -Costeño con Cuernos (CCC)- estimates of variance components for phenotypic and genetic parameters were obtained for birth weight, weight at weaning (adjusted to 270 days) and weight at 480 days. Using the restricted maximum likelihood (REMI) methodology, 2281 birth weight records (PN), 1722 weaning weight records and 1086 weight records adjusted to 480 days were analyze by fitting a model which included direct and maternal genetics effects as well as permanent environmental effects, assuming that fixed effects were year of birth weight, calf gender and the mother number of births. The genetic parameters for heritability, repeatability, genetic and phenotypic correlation were estimated and genotypic and phenotypic correlation was established. Heritability estimates for direct effects are low and range from 0.17 ± 0.001 and 0.21 ± 0.074 for birth and weaning weight respectively; while estimates for maternal genetics effects were also low for PN, they were higher for weaning weight and weight at 480 days. There was a negative correlation between direct and maternal effects, and the higher value was for PN (-0.89). The contribution of the variable permanent environment measured as the contribution of the phenotypic variance was low and diminished as animal age increased.</p>

scholarly journals Comparison of models for estimating genetic parameters and predicting breeding values for birth weight and calving ease in Czech Charolais cattle 

2014 ◽  
Vol 59 (No. 7) ◽  
pp. 302-309 ◽  
Author(s):  
L. Vostrý ◽  
Z. Veselá ◽  
A. Svitáková ◽  
H. Vostrá Vydrová
Keyword(s):  
Animal Model ◽  
Birth Weight ◽  
Beef Cattle ◽  
Fixed Effects ◽  
Genetic Parameters ◽  
Genetic Correlations ◽  
Genetic Effects ◽  
Parameters Estimation ◽  
Calving Ease ◽  
Heritability Estimates

The most appropriate model for genetic parameters estimation for calving ease and birth weight in beef cattle was selected. A total of 27 402 field records were available from the Czech Charolais breed. For estimation of genetic parameters for calving ease and body weight, three bivariate models were tested: a linear-linear animal model (L-LM) with calving ease classified into four categories (1 &ndash; easy; 2&ndash;4 &ndash; most difficult), a linear-linear animal model (SC-LM) in which calving ease scores were transformed into Snell scores (Snell 1964) and expressed as percentage of assisted calving (ranging 0&ndash;100%), and a bivariate threshold-linear animal model (T-LM) with calving ease classified into four categories (1 &ndash; easy, 2&ndash;4 &ndash; most difficult). All tested models included fixed effects for contemporary group (herd &times; year &times; season), age of dam, sex and breed of a calf. Random effects included direct and maternal genetic effects, maternal permanent environmental effect, and residual error. Direct heritability estimates for calving ease and birth weight were, with the use of L-LM, SC-LM, and T-LM, from 0.096 &plusmn; 0.013 to 0.226 &plusmn; 0.024 and from 0.210 &plusmn; 0.024 to 0.225 &plusmn; 0.026, respectively. Maternal heritability estimates for calving ease and birth weight were, with the use of L-LM, SC-LM, and T-LM, from 0.060 &plusmn; 0.031 to 0.104 &plusmn; 0.125 and from 0.074 &plusmn; 0.041 to 0.075 &plusmn; 0.040, respectively. Genetic correlations of direct calving ease with direct birth weight ranged from 0.46 &plusmn; 0.06 to 0.50 &plusmn; 0.06 for all tested models; whereas maternal genetic correlations between these two traits ranged from 0.24 &plusmn; 0.17 to 0.25 &plusmn; 0.53. Correlations between direct and maternal genetic effects within-trait were negative and substantial for all tested models (ranging from &ndash;0.574 &plusmn; 0.125 to &ndash;0.680 &plusmn; 0.141 for calving ease and from &ndash;0.553 &plusmn; 0.122 to &ndash;0.558 &plusmn; 0.118 for birth weight, respectively), illustrating the importance of including this parameter in calving ease evaluations. Results indicate that any of the tested models could be used to reliably estimate genetic parameters for calving ease for beef cattle in the Czech Republic. However, because of advantages in computation time and practical considerations, genetic analysis using SC-LM (transformed data) is recommended.

PHENOTYPIC AND GENETIC PARAMETER ESTIMATES FOR GESTATION LENGTH, CALVING EASE AND BIRTH WEIGHT IN SIMMENTAL CATTLE

1988 ◽  
Vol 68 (1) ◽  
pp. 291-294 ◽  
Author(s):  
R. A. KEMP ◽  
J. W. WILTON ◽  
L. R. SCHAEFFER
Keyword(s):  
Birth Weight ◽  
Random Mating ◽  
Genetic Parameter ◽  
Parameter Estimates ◽  
Gestation Length ◽  
Calving Ease ◽  
Phenotypic Correlations ◽  
Large Numbers ◽  
Genetic And Phenotypic Correlations ◽  
Variance And Covariance Components

Variance components, heritabilities and genetic and phenotypic correlations were estimated using progeny records of 73 sires on the Young Sire Proving Program (YSPP) of the Canadian Simmental Association. The YSPP was based on random mating of 58 test and 15 reference sires to cows in cooperating herds. Data were collected on several traits but the ones of interest in this study were gestation length (GL), calving ease (CE) and birth weight (BW). Variance and covariance components were estimated using Henderson's method 3. Heritability estimates were 0.24, 0.06 and 0.19 for GL, CE and BW, respectively. Genetic and phenotypic correlations were negative between GL and CE and CE and BW. Correlations between GL and BW were moderate and positive. Selection programs, utilizing large numbers of progeny per sire, would be effective but should incorporate the correlations between these traits. Key words: Heritabilities, genetic and phenotypic correlations, selection

Results of multivariate individual animal model genetic evaluations of british pedigree beef cattle

1997 ◽  
Vol 65 (2) ◽  
pp. 199-207 ◽  
Author(s):  
R. E. Crumps ◽  
G. Simm ◽  
D. Nicholson ◽  
R. H. Findlay ◽  
J. G. E. Bryan ◽  
...  
Keyword(s):  
Animal Model ◽  
Birth Weight ◽  
Beef Cattle ◽  
Fixed Effects ◽  
Breeding Value ◽  
Individual Animal ◽  
Group Sex ◽  
The Uk ◽  
Genetic Evaluations ◽  
Genetic Trends

AbstractThis paper reports the procedures put into place in the UK for the genetic evaluation of pedigree beef cattle and estimation of genetic trends using a comprehensive model to allow critical analysis of progress made under previous data recording schemes. Live weights of Simmental, Limousin, Charolais, South Devon and Aberdeen Angus beef cattle, recorded by the Meat and Livestock Commission (MLC) from 1970 to 1992 were analysed, as part of a project to introduce best linear unbiased predictions (BLUP) of breeding value in the British beef industry. Birth weights were available from MLC or the relevant breed society, (4000 to 84000 records, depending on the breed) and 200- and 400-day weights were estimated by within-animal linear regression on all available weights (resulting in 8000 to 48000 records per breed). Animals were retrospectively assigned to contemporary groups within herds, separately for each trait, taking account of observed calving patterns. Records were adjusted to correct for heterogeneity of variance between herds. BLUP evaluations were then performed within breed, fitting a multivariate individual animal model. In addition to additive direct genetic effects, additive maternal genetic and dam permanent environmental effects were included for birth weight and 200-day weight. Unknown parents were assigned to genetic groups, based on estimated date of birth. The model included fixed effects for contemporary group, sex, month of birth, birth type (single or multiple), embryo transfer births, fostered calves, breed of dam, proportion purebred and age of dam. Genetic trends were estimated by regressing estimated breeding values for animals on their year of birth. Trends in birth weight, 200-day weight and 400-day weight between 1970 and 1992 were approximately 0·09, 0·73 and 1·38 kg per annum respectively for the Charolais breed; 0·08, 0·76 and 1·33 kg per annum for the Simmental; 0·06, 0·53 and 0·89 kg per annum for the Limousin; 0·12, 1·02 and 1·86 kg per annum for the Aberdeen Angus; and 0·03, 0·38 and 0·82 kg per annum for the South Devon breed.

Sign in / Sign up

Export Citation Format

Share Document