Identification of key drought-tolerant genes in soybean using an integrative data-driven feature engineering pipeline

Soybean (Glycine max L.) is an important global food crop but is highly vulnerable to drought stress, which significantly impact agricultural productivity. Conventional methods for identifying key drought-tolerant genes (DTgenes), such as weighted gene co-expression network analysis (WGCNA), are resource-intensive, inefficient, and rely solely on gene expression data, limiting their capacity to integrate multi-dimensional datasets, reduce noise, and produce reliable outcomes. In this study, we propose a novel data-driven feature engineering (DFE) pipeline to address these limitations. The DFE pipeline integrates diverse genetic data from cloud-based sources, and applies systematic data extraction, transformation, loading, and integration of multi-dimensional omics and non-omics datasets. By employing robust feature prioritization and pathway-level functional evidence, it identifies biologically relevant candidate genes, minimizing false positives and enhancing discovery accuracy. The superiority of the DFE pipeline was demonstrated under conditions requiring integrative analysis of diverse datasets. Compared to WGCNA, the DFE pipeline demonstrated superior robustness to noise, and the ability to uncover biologically meaningful insights. The validation confirms the credibility of the identified key DTgenes in conferring drought tolerance response. Beyond improving efficiency and precision in soybean breeding programs, the DFE pipeline provides a scalable and adaptable framework for identifying critical genes in other crops. This innovation addresses the pressing challenges posed by climate change, offering a robust solution to enhance crop resilience and ensure sustainable agricultural productivity.