Refine
Has Fulltext
- yes (2)
Is part of the Bibliography
- yes (2)
Year of publication
- 2020 (2)
Document Type
- Journal article (1)
- Doctoral Thesis (1)
Language
- English (2)
Keywords
- GWAS (1)
- Genetics (1)
- Genomic Selection (1)
- Keras (1)
- Quantitative Genetics (1)
- breeding (1)
- deep learning (1)
- genomic selection (1)
- machine learning (1)
- phenotype prediction (1)
The prediction of breeding values and phenotypes is of central importance for both livestock and crop breeding. In this study, we analyze the use of artificial neural networks (ANN) and, in particular, local convolutional neural networks (LCNN) for genomic prediction, as a region-specific filter corresponds much better with our prior genetic knowledge on the genetic architecture of traits than traditional convolutional neural networks. Model performances are evaluated on a simulated maize data panel (n = 10,000; p = 34,595) and real Arabidopsis data (n = 2,039; p = 180,000) for a variety of traits based on their predictive ability. The baseline LCNN, containing one local convolutional layer (kernel size: 10) and two fully connected layers with 64 nodes each, is outperforming commonly proposed ANNs (multi layer perceptrons and convolutional neural networks) for basically all considered traits. For traits with high heritability and large training population as present in the simulated data, LCNN are even outperforming state-of-the-art methods like genomic best linear unbiased prediction (GBLUP), Bayesian models and extended GBLUP, indicated by an increase in predictive ability of up to 24%. However, for small training populations, these state-of-the-art methods outperform all considered ANNs. Nevertheless, the LCNN still outperforms all other considered ANNs by around 10%. Minor improvements to the tested baseline network architecture of the LCNN were obtained by increasing the kernel size and of reducing the stride, whereas the number of subsequent fully connected layers and their node sizes had neglectable impact. Although gains in predictive ability were obtained for large scale data sets by using LCNNs, the practical use of ANNs comes with additional problems, such as the need of genotyping all considered individuals, the lack of estimation of heritability and reliability. Furthermore, breeding values are additive by design, whereas ANN-based estimates are not. However, ANNs also comes with new opportunities, as networks can easily be extended to account for additional inputs (omics, weather etc.) and outputs (multi-trait models), and computing time increases linearly with the number of individuals. With advances in high-throughput phenotyping and cheaper genotyping, ANNs can become a valid alternative for genomic prediction.
Quantitative genetics is the study of continuously distributed traits and their ge-
netic components. Recent developments in DNA sequencing technologies and
computational systems allow researchers to conduct large scale in silico studies.
However, going from raw DNA reads to genomic prediction of quantitative traits
with the help of neural networks is a long and error-prone process. In the course
of this thesis, many steps involved in this process will be assessed in depth. Chap-
ter 2 will feature a study that compares the landscape of chloroplast genome as-
sembly tools. Chapter 3 will present a software to perform genome-wide associa-
tion studies using modern tools, which allow GWAS-Flow to outperform current
state of the art software packages. Chapter 4 will give an in depth introduc-
tion to machine learning and the nature of quantitative traits and will combine
those to genomic prediction with artificial neural networks and compares the re-
sults to those of algorithms based on linear mixed models. Finally, in Chapter 5
the results from the previous chapters are summarized and used to elucidate the
complex nature of studies concerning quantitative genetics.