DPCformer: An Interpretable Deep Learning Model for Genomic Prediction in Crops

Abstract

Translate Article

Addressing global food security demands more efficient crop breeding. While Genomic Selection (GS) accelerates breeding by predicting phenotypes from genomic data, its accuracy is limited by complex trait architectures and environmental dependencies. To overcome these challenges, we propose DPCformer, a deep learning model integrating convolutional neural networks with self-attention to capture nonlinear genotype-phenotype relationships. Evaluated on 13 traits across five crops (maize, cotton, tomato, rice, chickpea) using chromosomally-ordered SNP data processed with 8-dimensional encoding and PMF feature selection, DPCformer demonstrated superior performance. It achieved accuracy improvements of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.74-2.92 \%$</tex> in maize, up to 8.37% in cotton, and remarkable PCC enhancements of 57.35% in tomato and 16.62% in chickpea compared to baselines. These results establish DPCformer as a robust framework for genomic prediction, excelling in accuracy, small-sample performance, and polyploid data processing.