Predicting yield performance of parents in plant breeding: A neural collaborative filtering approach

Abstract

Experimental corn hybrids are created in plant breeding programs by crossing two parents, so-called inbred and tester, together. Identification of best parent combinations for crossing is challenging since the total number of possible cross combinations of parents is large and it is impractical to test all possible cross combinations due to limited resources of time and budget. In the 2020 Syngenta Crop Challenge, Syngenta released several large datasets that recorded the historical yield performances of around 4% of total cross combinations of 593 inbreds with 496 testers which were planted in 280 locations between 2016 and 2018 and asked participants to predict the yield performance of cross combinations of inbreds and testers that have not been planted based on the historical yield data collected from crossing other inbreds and testers. In this paper, we present a collaborative filtering method which is an ensemble of matrix factorization method and a neural network to solve this problem. Our computational results suggested that the proposed model significantly outperformed other models such as deep factorization machines (DeepFM), generalized matrix factorization (GMF), LASSO, random forest (RF), and neural networks. Presented method and results were produced within the 2020 Syngenta Crop Challenge.

A review of deep learning applications for genomic selection

“…); (d) there is much empirical evidence that the larger the dataset, the better the performance of DL models, which offers many opportunities to design specific topologies (deep neural networks) to deal with any type of data in a better way than current models used in GS, because DL models with topologies like CNN can very efficiently capture the correlation (special structure) between adjacent input variables, that is, linkage disequilibrium between nearby SNPs; (f) some DL topologies like CNN have the capability to significantly reduce the number of parameters (number of operations) that need to be estimated because CNN allows sharing parameters and performing data compression (using the pooling operation) without the need to estimate more parameters; and (g) the modeling paradigm of DL is closer to the complex systems that give rise to the observed phenotypic values of some traits. For these reasons, the incorporation of DL for classical breeding pipelines is in progress and some uses of DL are given next: 1) for the prediction of parental combinations, which is critical for choosing superior combinational homozygous parental lines in F1-hybrid breeding programs [ 84 ], 2) for modelling and predicting quantitative characteristics, for example, to perform image-based ear counting of wheat with high level of robustness, without considering variables, such as growth stage and weather conditions [ 85 ], 3) for genetic diversity and genotype classification, for example, in Cinnamomum osmophloeum Kanehira (Lauraceae), DL was applied to differentiate between morphologically similar species [ 86 ], and 4) for genomic selection (see Table 1 ).…”

Crop yield prediction using multi-parametric deep neural networks

“…However, in hybrid approach input weights reach near global minima at an early stage of iteration which leads to premature convergence. A neural collaborative filtering approach (8) was presented for predicting yield performance of parents in plant breeding. The prediction performance of this approach will be enhanced by including other important parameters such as weather components and soil conditions.…”

YieldNet: A Convolutional Neural Network for Simultaneous Corn and Soybean Yield Prediction Based on Remote Sensing Data

“…Deep learning models stack multiple non-linear modules which can learn very complex functions [21]. Deep learning methods have extensively been used to solve different problems in the field of agriculture such as crop yield prediction [22,23,24,25,26,27,28], image-based plant phenotyping [29,30,31,32], and drought tolerance classification [33]. These deep learning methods achieved high accuracy and improved the state-of-the-art in the field.…”