Abstract
Convolutional neural networks (CNN) are well suited image recognition tools for their ability to recognize latent patterns from images. Here, we investigated whether CNN can be used for genomic prediction. We created genomic images from genotype data and used them to predict phenotypes in mice. This approach was compared with traditional GBLUP and gradient boosting machine (GBM) models. For the two traits analysed, CNN was competitive in terms of predictive performance. The resolution of genomic images impacted model performance where, for this dataset, optimum results were obtained with 100×100 pixels. These first results demonstrate the potential of genomic images for genomic prediction using CNNs and merit investigation on adding layers of information to further increase accuracy of prediction.
Original language | English |
---|---|
Title of host publication | Proceedings of 12th World Congress on Genetics Applied to Livestock Production (WCGALP) |
Subtitle of host publication | Technical and species orientated innovations in animal breeding, and contribution of genetics to solving societal challenges |
Editors | R.F. Veerkamp, Y. de Haas |
Place of Publication | Wageningen |
Publisher | Wageningen Academic Publishers |
Pages | 1233-1236 |
ISBN (Electronic) | 9789086869404 |
DOIs | |
Publication status | Published - 2022 |
Event | World Congress on Genetics Applied to Livestock Production: WCGALP 2022 - Rotterdam, Netherlands Duration: 3 Jul 2022 → 8 Jul 2022 |
Conference/symposium
Conference/symposium | World Congress on Genetics Applied to Livestock Production: WCGALP 2022 |
---|---|
Country/Territory | Netherlands |
City | Rotterdam |
Period | 3/07/22 → 8/07/22 |