Legume Genomics and Genetics 2025, Vol.16, No.1, 44-53 http://cropscipublisher.com/index.php/lgg 50 truly cultivate high-quality varieties. Just like someone specifically focuses on high-protein sites for breeding (Qin et al., 2022), the soybeans cultivated in this way are more suitable for use as feed or processed foods. Breeding nowadays is much smarter than before. Applying these markers found in GWAS to MAS or GS (Stewart-Brown et al., 2019) can not only address climate change but also improve nutritional quality. But to be honest, although these technologies are powerful, ultimately it still depends on how the soybeans grow in the fields are (McLeod et al., 2023). After all, no matter how good the genes are, they must stand the test of actual cultivation. 6 Limitations and Future Directions 6.1 Limitations of GWAS (population size, environmental interactions) GWAS studies on soybean genes have indeed achieved considerable success over the years, but there are also many problems in actual operation. Take the sample size for example. The population size used in many studies is really too small - for instance, one study found some key sites using 304 soybean strains (Sonah et al., 2015), but the researchers themselves admitted that the sample size would have to be increased to make the results more reliable (Shook et al., 2021). This is like conducting a public opinion poll. If too few people are asked, the result is bound to be unreliable. What's more troublesome is that the important traits of soybeans, such as yield and plant height, are constantly affected by the environment. It's often hard to tell whether it's the genes at play or the weather causing trouble. Fortunately, there is now a meta-GWAS method that can analyze data from multiple studies together. The conclusions drawn in this way are indeed more convincing. Ultimately, no matter how powerful GWAS is, it is still a tool. The key lies in how people use it. 6.2 Integration with other genomic tools (QTL mapping, CRISPR) Although GWAS is useful, it is really not enough to rely on it alone. Nowadays, when conducting research, a combination of measures is emphasized - for instance, incorporating QTL mapping as an assistant. Just like those loci found in previous GWAS, when re-validated in the parent population using QTL (Sonah et al., 2015), the results were indeed much more reliable. However, the most remarkable one is the CRISPR technology that has become popular in the past two years. It is simply a "microscope" for genetic research (Contreras-Soto et al., 2017). Want to confirm which gene is useful? Just edit it out and see the effect (Kim et al., 2022). Although each of these technologies has its own tricks, when used together, they can indeed complement each other's shortcomings. To put it bluntly, nowadays, when studying genes, one not only needs to be able to "find" them but also "verify" them, so as to truly contribute to breeding work. 6.3 Opportunities for multi-trait GWAS and meta-analysis There is a new trend in the study of soybean genes now - the joint efforts of multi-trait GWAS and meta-analysis. This trick is quite interesting. It can identify the "versatile" genes that affect multiple traits at once. For instance, one study aggregated data from 73 independent experiments (Shook et al., 2021), and as a result, 483 QTLS were identified, many of which were associated with several agronomic traits simultaneously. More practical is the meta-analysis method, which involves analyzing research data from different teams and in different environments together (Hu et al., 2021). The gene loci found in this way are particularly reliable. Key sites like Joukhadar et al. (2021) that can affect yield and plant height regardless of weather changes were identified in this way. However, to be honest, although these new methods are very powerful, they can be quite mentally challenging to analyze when the volume of data grows large. 6.4 Future trends in soybean genomics research The research on soybean genes may be going to play some new tricks next. Those algorithms in machine learning are quite popular now, such as support vector regression (SVR) and random forest (RF) (Yoosefzadeh-Najafabadi et al., 2021) It is indeed more accurate and faster to find QTL in GWAS (Yoosefzadeh-Najafabadi et al., 2023). Sequencing technology is also becoming increasingly advanced. It is estimated that in the future, when analyzing the soybean genome, we will be able to see it more thoroughly. However, the most urgent matter might still be the study of climate adaptability - after all, the weather is getting more and more strange now, and it is necessary to
RkJQdWJsaXNoZXIy MjQ4ODYzNA==