Legume Genomics and Genetics 2025, Vol.16, No.1, 11-22 http://cropscipublisher.com/index.php/lgg 19 When it comes to drought resistance of soybeans, nowadays research has to adopt a "combination punch" (Wang et al., 2020). Just looking at the genome is not enough; it is also necessary to analyze the data of the transcriptome, proteome, and metabolome together-just like having to read a score, listen to a performance, and study the movements of each musician. Although this approach is troublesome, it can indeed reveal quite a few tricks: for instance, when will a certain drought-resistant gene be activated? Which proteins will be driven to work? What metabolites were finally produced? Only when these clues are strung together can a complete drought resistance mechanism be pieced together. To be honest, with so much omics data piled up together, just processing it is enough to make a big deal out of it. The existing analysis tools often get stuck. What is most needed now is smarter algorithms that can automatically correlate data at different levels and, ideally, predict the regulatory relationships between genes. Although there is still some way to go before the drought-resistant network is completely cracked, at least the path is becoming clearer and clearer as we go along. 7.3 In-depth exploration of gene-environment interaction studies When it comes to soybeans' drought resistance, it's not something that can be decided by genes alone-environmental factors are the real "masterminds" behind the scenes. The research in 2022 was particularly interesting (Sun et al., 2022). The performance of the same drought-resistant genes in sandy soil and yellow clay can be worlds apart. As soon as the soil moisture changes, the gene expression pattern seems to undergo a magic trick. The temperature level is directly related to the effectiveness of drought resistance. Sometimes, genes that work in high temperatures can be counterproductive in low temperatures. The most infuriating thing is that nutritional status also gets involved. Different levels of nitrogen, phosphorus and potassium result in different "work enthusiasm" of genes. Although QTL mapping is doing very well now, we still know too little about how these genes "tailor their treatment to the individual" in different environments. Just like the same actor can perform completely different effects in different film crews, the ins and outs of this are enough for researchers to ponder over for another ten or eight years. Indeed, to truly crack the code for soybeans' drought resistance, merely tinkering in the laboratory is not enough. Recently, some scholars have proposed (Wang and Li., 2024) that the same materials should be planted all over the country, from the black soil of Northeast China to the arid areas of Northwest China, so as to screen out those reliable QTLS that "shine brightly with a little sunlight". It's easier said than done. Such large-scale experiments are both costly and time-consuming. But there's no way around it. After all, genes are just like people; a different environment might completely change their appearance. In 2016, a study pointed out (Chang et al., 2016) that the current gene-environment interaction model is still too "rigid" and a more intelligent prediction system needs to be developed. Imagine if, like weather forecasts, parameters such as soil moisture and temperature could be input to predict the expression of a certain genotype in a specific area. Then, breeders would have a much easier time. However, to be fair, the way soybeans interact with the environment is just too complex. If this system is really to be developed, it will probably take several more years of hard work. 7.4 Application of big data and machine learning in GWAS research Big data and machine learning have now become the "golden pair" in GWAS research-this is particularly evident when analyzing the drought resistance traits of soybeans. Just think about it. GWAS often generates massive amounts of genotype and phenotype data (Zhang et al., 2021), and traditional statistical methods handle them as if calculating space orbits. But machine learning is different, especially deep learning, which can dig out hidden patterns from these data that the human eye cannot detect at all. For instance, in 2021, a team successfully experimented with other crops. The predictive model they developed could even capture those nonlinear gene interactions. To be honest, however, there are still many problems to be solved when applying these techniques to the research on soybean drought resistance: the data quality varies greatly, the model is prone to overfitting, and the most troublesome issue is the explainability of the results. However, in any case, this approach has indeed given researchers new hope. At least they no longer have to stare blankly at piles of data.
RkJQdWJsaXNoZXIy MjQ4ODYzNA==