{
"resultCount": 4,
"records": [
{
"id": "UFV_2fd841b63a89d1682d400371a8c8e782",
"title": "Redes neurais artificiais na discrimina\u00e7\u00e3o de popula\u00e7\u00f5es de retrocruzamento com diferentes graus de similaridade",
"abstract_por": [
"A correta classifica\u00e7\u00e3o de indiv\u00edduos \u00e9 de extrema import\u00e2ncia para fins de preserva\u00e7\u00e3o da variabilidade gen\u00e9tica existente bem como para a maximiza\u00e7\u00e3o dos ganhos. As t\u00e9cnicas de estat\u00edstica multivariada comumente utilizada nessas situa\u00e7\u00f5es s\u00e3o as fun\u00e7\u00f5es discriminantes de Fisher e de Anderson, que permitem alocar um indiv\u00edduo inicialmente desconhecido em uma das g popula\u00e7\u00f5es prov\u00e1veis ou grupos pr\u00e9-definidos. Entretanto, para altos n\u00edveis de similaridade como \u00e9 o caso de popula\u00e7\u00f5es de retrocruzamentos esses m\u00e9todos tem se mostrado pouco eficientes. Atualmente, muito se fala de um novo paradigma de computa\u00e7\u00e3o, as redes neurais artificiais, que podem ser utilizadas para resolver diversos problemas da Estat\u00edstica, como agrupamento de indiv\u00edduos similares, previs\u00e3o de s\u00e9ries temporais e em especial, os problemas de classifica\u00e7\u00e3o. O objetivo desse trabalho foi realizar um estudo comparativo entre as fun\u00e7\u00f5es discriminantes de Fisher e de Anderson e as redes neurais artificiais quanto ao n\u00famero de classifica\u00e7\u00f5es incorretas de indiv\u00edduos sabidamente pertencentes a diferentes popula\u00e7\u00f5es simuladas de retrocruzamento, com crescentes n\u00edveis de similaridade. A dissimilaridade, medida pela dist\u00e2ncia de Mahalanobis, foi um conceito de fundamental import\u00e2ncia na utiliza\u00e7\u00e3o das t\u00e9cnicas de discrimina\u00e7\u00e3o, pois quantificou o quanto as popula\u00e7\u00f5es eram divergentes. A obten\u00e7\u00e3o dos dados foi feita atrav\u00e9s de simula\u00e7\u00e3o utilizando o programa computacional Genes. Cada popula\u00e7\u00e3o, gerada por simula\u00e7\u00e3o, foi caracterizada por um conjunto de elementos mensurados por caracter\u00edsticas de natureza cont\u00ednua. Foram geradas considerados 50 locos independentes, cada qual com dois alelos. As rela\u00e7\u00f5es de parentescos e a estrutura\u00e7\u00e3o hier\u00e1rquica foram estabelecidas considerando popula\u00e7\u00f5es genitoras geneticamente divergentes, h\u00edbrido F1 e cinco gera\u00e7\u00f5es de retrocruzamento em rela\u00e7\u00e3o a cada um dos genitores, permitindo estabelecer par\u00e2metros de efic\u00e1cia das metodologias testadas. Os dados fenot\u00edpicos das popula\u00e7\u00f5es foram utilizados para estabelecimento da fun\u00e7\u00e3o discriminante de Fisher e Anderson e para o c\u00e1lculo da taxa de erro aparente (TEA), que mede o xi n\u00famero de classifica\u00e7\u00f5es incorretas. As estimativas de TEA foram comparadas com as obtida por meio das Redes Neurais Artificiais. As redes neurais artificiais mostraram-se uma t\u00e9cnica promissora no que diz respeito a problemas de classifica\u00e7\u00e3o, uma vez que apresentaram um n\u00famero de classifica\u00e7\u00f5es incorretas de indiv\u00edduos menor que os dados obtidos pelas fun\u00e7\u00f5es discriminantes."
],
"abstract_eng": [
"The correct classification of individuals has a top importance for the genetic variability preservation as well as to maximize gains. The multivariate statistical techniques commonly used in these situations are the Fisher and Anderson discriminant functions, allowing to allocate an initially unknown individual in a probably g population or predefined groups. However, for higher levels of similarity such as backcross populations these methods has proved to be inefficient. Currently, much has been Said about a new paradigm of computing, artificial neural networks, which can be used to solve many statistical problems as similar subjects grouping, time-series forecasting and in particular, the classification problems. The aim of this study was to conduct a comparative study between the Fisher and Anderson discriminant functions and artificial neural networks through the number of incorrect classifications of individuals known to belong to different simulated backcross with increasing levels of populations similarity. The dissimilarity, measured by Mahalanobis distance, was a concept of fundamental importance in the use of discrimination techniques, due the quantification of how much populations were divergent. Data collection was done through simulation using the software Genes. Each population generated was characterized by a set of elements measured by characteristics of a continuous distribution. The relations of relatives and hierarchical structuring were established considering genetically divergent populations, F1 hybrid and five generations of backcrossing in relation to each of the relatives, establishing measures of effectiveness of the tested methodologies. The phenotypic data of populations were used to establish the Fisher and Anderson discriminant function and the calculation of the apparent error rate (AER), which measures the number of incorrect classifications. The ERA Estimations were compared with those obtained by means of neural networks. The artificial neural network is shown as a promising technique to solve classification problems, once it had a number of incorrect individuals classifications smaller than the data obtained by the discriminant functions."
],
"authors": {
"primary": {
"Sant'anna, Isabela de Castro": {
"profile": [
"http:\/\/lattes.cnpq.br\/0822371511052579"
]
}
}
},
"contributors": {
"advisor": {
"Cruz, Cosme Dami\u00e3o": {
"profile": [
"http:\/\/buscatextual.cnpq.br\/buscatextual\/visualizacv.do?id=K4788274A6"
]
}
},
"coadvisor": {
"Bhering, Leonardo Lopes": {
"profile": [
"http:\/\/buscatextual.cnpq.br\/buscatextual\/visualizacv.do?id=K4764363E6"
]
},
"Carneiro, Pedro Cresc\u00eancio Souza": {
"profile": [
"http:\/\/buscatextual.cnpq.br\/buscatextual\/visualizacv.do?id=K4728227T6"
]
}
},
"referee": {
"Nascimento, Moys\u00e9s": {
"profile": [
"http:\/\/lattes.cnpq.br\/6544887498494945"
]
}
}
},
"subjectsCNPQ": [
[
"CNPQ::CIENCIAS BIOLOGICAS::GENETICA::GENETICA QUANTITATIVA"
]
],
"subjectsPOR": [
[
"Melhoramento gen\u00e9tico"
],
[
"An\u00e1lise discriminante"
],
[
"Intelig\u00eancia artificial"
],
[
"Redes neurais"
]
],
"subjectsENG": [
[
"Breeding"
],
[
"Discriminant analysis"
],
[
"Artificial intelligence"
],
[
"Neural networks"
]
],
"institutions": [
"UFV"
],
"departements": [
"Gen\u00e9tica animal; Gen\u00e9tica molecular e de microrganismos; Gen\u00e9tica quantitativa; Gen\u00e9tica vegetal; Me"
],
"programs": [
"Mestrado em Gen\u00e9tica e Melhoramento"
],
"types": [
"masterThesis"
],
"accesslevel": "openAccess",
"publicationDates": [
"2014"
],
"urls": [
"http:\/\/locus.ufv.br\/handle\/123456789\/4801"
],
"formats": [
"masterThesis"
],
"languages": [
"por"
]
},
{
"id": "UFV_6771dde4d3e3deed6118985064b9ad95",
"title": "Redes neurais artificiais para predi\u00e7\u00e3o gen\u00f4mica na presen\u00e7a de intera\u00e7\u00f5es epist\u00e1ticas",
"abstract_eng": [
"The identification of elite individual is a critical component of most plant breeding programs. However, the ability to achieve this goal is limited by the high cost of phenotyping and conducting experiments. In this context the genomic selection was proposed to use all marks presents in the genome to estimate the genomic breeding value of individuals (GEBV) without the need to phenotyping. However, most applications of GS includes only the additive portion of the genetic value, and a more realistic representation of the genetic architecture of quantitative traits should have the inclusion of dominance and epistatics interaction. The role of epistasis in the genetic architecture of quantitative traits has been debated since first formulations of quantitative genetic theory, and different perspectives regarding the importance of epistasis arise. In populations, the total genetic variance is partitioned into components that are attributable to additive, dominance and epistatic variance, which depend on allele frequencies. If the allele frequency of the interacting locus varies among populations, the effect of the target locus can be significant in one population but not in another, or can even be of the opposite sign. In this context, Artificial Neural Networks (ANNs) has a great potential because they can capture non-linear relationships between markers from the data themselves, which most of the models commonly used in the GS can not. However, the inclusion of all markers in the prediction model increases the chances of a high correlation between the marks and represents a huge challenge that add less precision and a great computational demand for ANNs training that use a good part of their resources to represent irrelevant portions of the search space and compromising the learning process. Thus, a more realistic model should include only SNPs that are related to the traits of interest. Because of this, it was proposed to use dimensionality reduction methods, applied to the prediction of genetic values, for the purpose of selecting a subset of markers by means of specific procedures such as Sonda or Stepwise regressions. In this way, the objective of this work is to evaluate the efficiency of genome enabled prediction by using RR-BLUP (GS) and artificial neural networks as radial basis function neural network (RBFNN), and Multi-layer Perceptron (RNA-MLP) in the prediction of the genetic value in a natural population with linkage disequilibrium without (chapter 1) and with (chapter 2) the dimensionality reduction. For this, an Fl population from the hybridization of divergent parents with 500 individuals genotyped with l,000 SNP-type markers was simulated. The phenotypic traits were determined by adopting three different gene action models: additive, additive-dominance and epistasis, attending two dominance situations: partial and complete with quantitative traits admitting heritabilities (hz) ranging from 30 to 60%, each is controlled by 50 loci, considering two alleles per loco, totaling 12 different scenarios. To evaluate the predictive ability of RR-BLUP and the neural networks a cross- validation procedure with five replicates were trained using 80% of the individuals of the population. Two dimensionality reduction methods Stepwise and Sonda were used to calculated the square of the correlation between predicted genomic value (GEBV) and genotype\/phenotype value was used to measure predictive reliability(R2) and the predictive mean-squared error root (MSER). In the chapter one of this work the results showed that the use of neural networks allows capturing the epistasic interactions leading to an improvement in the accuracy of the prediction of the genetic value and, mainly, a large reduction of the mean square error root (MSER) that indicates greater reliability of the prediction of the genomic value. But from the results using phenotypic validation it was clearly that is possible to make further improvements on the accuracy by introducing the variable selection. Consequently, in the second chapter, after applied the dimensionality reduction methods, the the accuracy increased. For example, for h2 = 0.3 in the additive scenario, the validation R2 was 59% for neural network (RBFNN), 57% (RNA-MLP) and 57% for RR-BLUP, and in the epistemic scenario R2 values were 50%, 47 and 41%, respectively. Additionally, when analyzing the mean-squared error root the difference in performance of the techniques is even greater. For additive scenario, the estimates were 9l (RR-BLUP) and 5 for both neural networks and, in the most critical scenario, 427 (RR-BLUP) and 20 for neural networks. The results show that the use of neural networks allows capturing the epistasis interactions leading to an improvement in the accuracy of the prediction of the genetic value and, mainly, a large reduction of the mean square error root that indicates greater reliability of the prediction of the genomic value."
],
"authors": {
"primary": {
"Sant'anna, Isabela de Castro": {
"profile": [
"http:\/\/lattes.cnpq.br\/0822371511052579"
]
}
}
},
"contributors": {
"advisor": {
"Cruz, Cosme Dami\u00e3o": {
"profile": [
[
"NA"
]
]
}
}
},
"subjectsCNPQ": [
[
"Gen\u00e9tica Quantitativa"
]
],
"institutions": [
"UFV"
],
"types": [
"doctoralThesis"
],
"accesslevel": "openAccess",
"publicationDates": [
"2018"
],
"urls": [
"http:\/\/www.locus.ufv.br\/handle\/123456789\/20126"
],
"formats": [
"doctoralThesis"
],
"languages": [
"por"
]
},
{
"id": "UFV_d9b3eb0c83e7c903f737ef772577e4aa",
"title": "Redes neurais artificiais para predi\u00e7\u00e3o gen\u00f4mica na presen\u00e7a de intera\u00e7\u00f5es epist\u00e1ticas",
"abstract_eng": [
"The identification of elite individual is a critical component of most plant breeding programs. However, the ability to achieve this goal is limited by the high cost of phenotyping and conducting experiments. In this context the genomic selection was proposed to use all marks presents in the genome to estimate the genomic breeding value of individuals (GEBV) without the need to phenotyping. However, most applications of GS includes only the additive portion of the genetic value, and a more realistic representation of the genetic architecture of quantitative traits should have the inclusion of dominance and epistatics interaction. The role of epistasis in the genetic architecture of quantitative traits has been debated since first formulations of quantitative genetic theory, and different perspectives regarding the importance of epistasis arise. In populations, the total genetic variance is partitioned into components that are attributable to additive, dominance and epistatic variance, which depend on allele frequencies. If the allele frequency of the interacting locus varies among populations, the effect of the target locus can be significant in one population but not in another, or can even be of the opposite sign. In this context, Artificial Neural Networks (ANNs) has a great potential because they can capture non-linear relationships between markers from the data themselves, which most of the models commonly used in the GS can not. However, the inclusion of all markers in the prediction model increases the chances of a high correlation between the marks and represents a huge challenge that add less precision and a great computational demand for ANNs training that use a good part of their resources to represent irrelevant portions of the search space and compromising the learning process. Thus, a more realistic model should include only SNPs that are related to the traits of interest. Because of this, it was proposed to use dimensionality reduction methods, applied to the prediction of genetic values, for the purpose of selecting a subset of markers by means of specific procedures such as Sonda or Stepwise regressions. In this way, the objective of this work is to evaluate the efficiency of genome enabled prediction by using RR-BLUP (GS) and artificial neural networks as radial basis function neural network (RBFNN), and Multi-layer Perceptron (RNA-MLP) in the prediction of the genetic value in a natural population with linkage disequilibrium without (chapter 1) and with (chapter 2) the dimensionality reduction. For this, an Fl population from the hybridization of divergent parents with 500 individuals genotyped with l,000 SNP-type markers was simulated. The phenotypic traits were determined by adopting three different gene action models: additive, additive-dominance and epistasis, attending two dominance situations: partial and complete with quantitative traits admitting heritabilities (hz) ranging from 30 to 60%, each is controlled by 50 loci, considering two alleles per loco, totaling 12 different scenarios. To evaluate the predictive ability of RR-BLUP and the neural networks a cross- validation procedure with five replicates were trained using 80% of the individuals of the population. Two dimensionality reduction methods Stepwise and Sonda were used to calculated the square of the correlation between predicted genomic value (GEBV) and genotype\/phenotype value was used to measure predictive reliability(R2) and the predictive mean-squared error root (MSER). In the chapter one of this work the results showed that the use of neural networks allows capturing the epistasic interactions leading to an improvement in the accuracy of the prediction of the genetic value and, mainly, a large reduction of the mean square error root (MSER) that indicates greater reliability of the prediction of the genomic value. But from the results using phenotypic validation it was clearly that is possible to make further improvements on the accuracy by introducing the variable selection. Consequently, in the second chapter, after applied the dimensionality reduction methods, the the accuracy increased. For example, for h2 = 0.3 in the additive scenario, the validation R2 was 59% for neural network (RBFNN), 57% (RNA-MLP) and 57% for RR-BLUP, and in the epistemic scenario R2 values were 50%, 47 and 41%, respectively. Additionally, when analyzing the mean-squared error root the difference in performance of the techniques is even greater. For additive scenario, the estimates were 9l (RR-BLUP) and 5 for both neural networks and, in the most critical scenario, 427 (RR-BLUP) and 20 for neural networks. The results show that the use of neural networks allows capturing the epistasis interactions leading to an improvement in the accuracy of the prediction of the genetic value and, mainly, a large reduction of the mean square error root that indicates greater reliability of the prediction of the genomic value."
],
"authors": {
"primary": {
"Sant'anna, Isabela de Castro": {
"profile": [
"http:\/\/lattes.cnpq.br\/0822371511052579"
]
}
}
},
"contributors": {
"advisor": {
"Cruz, Cosme Dami\u00e3o": {
"profile": [
[
"NA"
]
]
}
}
},
"subjectsCNPQ": [
[
"Gen\u00e9tica Quantitativa"
]
],
"institutions": [
"UFV"
],
"types": [
"doctoralThesis"
],
"accesslevel": "openAccess",
"publicationDates": [
"2018"
],
"urls": [
"http:\/\/www.locus.ufv.br\/handle\/123456789\/20126"
],
"formats": [
"doctoralThesis"
],
"languages": [
"por"
]
},
{
"id": "UFV_5043f214c9a070214ee184182c25a656",
"title": "Redes neurais artificiais na discrimina\u00e7\u00e3o de popula\u00e7\u00f5es de retrocruzamento com diferentes graus de similaridade",
"abstract_por": [
"A correta classifica\u00e7\u00e3o de indiv\u00edduos \u00e9 de extrema import\u00e2ncia para fins de preserva\u00e7\u00e3o da variabilidade gen\u00e9tica existente bem como para a maximiza\u00e7\u00e3o dos ganhos. As t\u00e9cnicas de estat\u00edstica multivariada comumente utilizada nessas situa\u00e7\u00f5es s\u00e3o as fun\u00e7\u00f5es discriminantes de Fisher e de Anderson, que permitem alocar um indiv\u00edduo inicialmente desconhecido em uma das g popula\u00e7\u00f5es prov\u00e1veis ou grupos pr\u00e9-definidos. Entretanto, para altos n\u00edveis de similaridade como \u00e9 o caso de popula\u00e7\u00f5es de retrocruzamentos esses m\u00e9todos tem se mostrado pouco eficientes. Atualmente, muito se fala de um novo paradigma de computa\u00e7\u00e3o, as redes neurais artificiais, que podem ser utilizadas para resolver diversos problemas da Estat\u00edstica, como agrupamento de indiv\u00edduos similares, previs\u00e3o de s\u00e9ries temporais e em especial, os problemas de classifica\u00e7\u00e3o. O objetivo desse trabalho foi realizar um estudo comparativo entre as fun\u00e7\u00f5es discriminantes de Fisher e de Anderson e as redes neurais artificiais quanto ao n\u00famero de classifica\u00e7\u00f5es incorretas de indiv\u00edduos sabidamente pertencentes a diferentes popula\u00e7\u00f5es simuladas de retrocruzamento, com crescentes n\u00edveis de similaridade. A dissimilaridade, medida pela dist\u00e2ncia de Mahalanobis, foi um conceito de fundamental import\u00e2ncia na utiliza\u00e7\u00e3o das t\u00e9cnicas de discrimina\u00e7\u00e3o, pois quantificou o quanto as popula\u00e7\u00f5es eram divergentes. A obten\u00e7\u00e3o dos dados foi feita atrav\u00e9s de simula\u00e7\u00e3o utilizando o programa computacional Genes. Cada popula\u00e7\u00e3o, gerada por simula\u00e7\u00e3o, foi caracterizada por um conjunto de elementos mensurados por caracter\u00edsticas de natureza cont\u00ednua. Foram geradas considerados 50 locos independentes, cada qual com dois alelos. As rela\u00e7\u00f5es de parentescos e a estrutura\u00e7\u00e3o hier\u00e1rquica foram estabelecidas considerando popula\u00e7\u00f5es genitoras geneticamente divergentes, h\u00edbrido F1 e cinco gera\u00e7\u00f5es de retrocruzamento em rela\u00e7\u00e3o a cada um dos genitores, permitindo estabelecer par\u00e2metros de efic\u00e1cia das metodologias testadas. Os dados fenot\u00edpicos das popula\u00e7\u00f5es foram utilizados para estabelecimento da fun\u00e7\u00e3o discriminante de Fisher e Anderson e para o c\u00e1lculo da taxa de erro aparente (TEA), que mede o xi n\u00famero de classifica\u00e7\u00f5es incorretas. As estimativas de TEA foram comparadas com as obtida por meio das Redes Neurais Artificiais. As redes neurais artificiais mostraram-se uma t\u00e9cnica promissora no que diz respeito a problemas de classifica\u00e7\u00e3o, uma vez que apresentaram um n\u00famero de classifica\u00e7\u00f5es incorretas de indiv\u00edduos menor que os dados obtidos pelas fun\u00e7\u00f5es discriminantes."
],
"abstract_eng": [
"The correct classification of individuals has a top importance for the genetic variability preservation as well as to maximize gains. The multivariate statistical techniques commonly used in these situations are the Fisher and Anderson discriminant functions, allowing to allocate an initially unknown individual in a probably g population or predefined groups. However, for higher levels of similarity such as backcross populations these methods has proved to be inefficient. Currently, much has been Said about a new paradigm of computing, artificial neural networks, which can be used to solve many statistical problems as similar subjects grouping, time-series forecasting and in particular, the classification problems. The aim of this study was to conduct a comparative study between the Fisher and Anderson discriminant functions and artificial neural networks through the number of incorrect classifications of individuals known to belong to different simulated backcross with increasing levels of populations similarity. The dissimilarity, measured by Mahalanobis distance, was a concept of fundamental importance in the use of discrimination techniques, due the quantification of how much populations were divergent. Data collection was done through simulation using the software Genes. Each population generated was characterized by a set of elements measured by characteristics of a continuous distribution. The relations of relatives and hierarchical structuring were established considering genetically divergent populations, F1 hybrid and five generations of backcrossing in relation to each of the relatives, establishing measures of effectiveness of the tested methodologies. The phenotypic data of populations were used to establish the Fisher and Anderson discriminant function and the calculation of the apparent error rate (AER), which measures the number of incorrect classifications. The ERA Estimations were compared with those obtained by means of neural networks. The artificial neural network is shown as a promising technique to solve classification problems, once it had a number of incorrect individuals classifications smaller than the data obtained by the discriminant functions."
],
"authors": {
"primary": {
"Sant'anna, Isabela de Castro": {
"profile": [
"http:\/\/lattes.cnpq.br\/0822371511052579"
]
}
}
},
"contributors": {
"advisor": {
"Cruz, Cosme Dami\u00e3o": {
"profile": [
"http:\/\/buscatextual.cnpq.br\/buscatextual\/visualizacv.do?id=K4788274A6"
]
}
},
"coadvisor": {
"Bhering, Leonardo Lopes": {
"profile": [
"http:\/\/buscatextual.cnpq.br\/buscatextual\/visualizacv.do?id=K4764363E6"
]
},
"Carneiro, Pedro Cresc\u00eancio Souza": {
"profile": [
"http:\/\/buscatextual.cnpq.br\/buscatextual\/visualizacv.do?id=K4728227T6"
]
}
},
"referee": {
"Nascimento, Moys\u00e9s": {
"profile": [
"http:\/\/lattes.cnpq.br\/6544887498494945"
]
}
}
},
"subjectsCNPQ": [
[
"CNPQ::CIENCIAS BIOLOGICAS::GENETICA::GENETICA QUANTITATIVA"
]
],
"subjectsPOR": [
[
"Melhoramento gen\u00e9tico"
],
[
"An\u00e1lise discriminante"
],
[
"Intelig\u00eancia artificial"
],
[
"Redes neurais"
]
],
"subjectsENG": [
[
"Breeding"
],
[
"Discriminant analysis"
],
[
"Artificial intelligence"
],
[
"Neural networks"
]
],
"institutions": [
"UFV"
],
"departements": [
"Gen\u00e9tica animal; Gen\u00e9tica molecular e de microrganismos; Gen\u00e9tica quantitativa; Gen\u00e9tica vegetal; Me"
],
"programs": [
"Mestrado em Gen\u00e9tica e Melhoramento"
],
"types": [
"masterThesis"
],
"accesslevel": "openAccess",
"publicationDates": [
"2014"
],
"urls": [
"http:\/\/locus.ufv.br\/handle\/123456789\/4801"
],
"formats": [
"masterThesis"
],
"languages": [
"por"
]
}
],
"status": "OK"
}