DNA Based Crop Selector System Using AI

Genomic Selection (GS) utilizing deep learning offers significant potential to accelerate crop breeding. However, state-of-the-art interpretable deep learning frameworks, often combining Convolutional Neural Networks (CNNs) and multi-head self-attention, frequently exhibit limited real-world applicability due to their exclusion of environmental factors, failing to capture crucial Genotype x Environment (GxE) interactions. This research addresses this limitation by enhancing such an interpretable framework through the integration of key environmental variables. We collected and engineered relevant weather (NASA POWER) and soil (ISRIC SoilGrids) data corresponding to the datasets of five major crops: maize, rice, wheat, foxtail millet, and tomato, leveraging publicly available datasets used in foundational studies. By incorporating these environmental features alongside the genomic (SNP) data within the deep learning architecture, we developed GxE-aware prediction models. This work demonstrates a comprehensive methodology for building GxE-enabled deep learning models for genomic prediction across diverse crop species, aiming to improve predictive accuracy and provide a more robust tool for practical plant breeding applications.