A Gated Convolutional Neural Network-Based Method for Crop Phenotype Prediction

Genomic selection (GS) is a modern breeding technology that predicts the phenotypic performance of crops or animals based on genomic data, significantly improving breeding efficiency. Due to the complex relationship between genotype and phenotype, traditional methods have shown limited effectiveness for certain crops. In recent years, deep learning has emerged as a research hotspot in the GS field due to its ability to capture nonlinear relationships. In this paper, we propose a method for predicting crop phenotypes based on gated convolutional neural networks, named GatedConv. GatedConv consists of two gated convolution blocks and two fully connected layers. The gating mechanism within the gated convolutional blocks controls the activation level of the convolutional outputs, thereby filtering redundant features. We applied GatedConv to genotype and phenotype data from wheat, soybean, and rice, and compared its accuracy with other state-of-the-art GS methods. Experimental results demonstrate that the proposed GatedConv exhibits superior performance.