Enhanced Genomic Prediction for Maize Breeding Using Deep Learning and K-mer-based Sequence Encoding

Maize (Zea mays L.) plays a pivotal role in global food security due to its substantial economic and nutritional value. With advancements in agricultural technologies, crop yields have surged, but challenges remain in optimizing maize breeding for even higher productivity. This research explores cutting-edge approaches in maize breeding, focusing on novel genomic prediction tools that enhance crop yield. While numerous studies have demonstrated the potential of genomic prediction in accelerating maize yields, current models often suffer from suboptimal data quality and lengthy processing times due to the use of raw data for genotype prediction. To address these challenges, this study introduces a deep learning- based model that leverages polymorphism data, employs k-mer methods to shorten sequences, and applies one-hot encoding for preprocessing. This data is then processed by a convolutional neural network (CNN) to predict genotypes efficiently. The proposed model aims to achieve a 98% accuracy in genomic prediction, outperforming existing models in both speed and accuracy, thereby contributing to the generation of superior maize traits.