Researchers at Huazhong Agricultural University (HZAU) have unveiled Fast3VmrMLM, a cutting-edge algorithm that integrates genome-wide scanning with machine learning, significantly boosting the efficiency of gene discovery and crop breeding for complex polygenic traits. The research, led by professor Zhang Yuanming's team, was recently published in Plant Communications.

Fast3VmrMLM represents a major leap forward in genome-wide association studies (GWAS), as it is capable of handling massive datasets at reduced computational costs and time. The algorithm shortens conventional GWAS runtime from hours to minutes, even when analyzing populations of 500,000 varieties with over 100 million markers — on a relatively modest server (20 CPUs, 1 TB RAM).

By combining seven innovative techniques, the algorithm addresses key limitations in traditional association methods. It enhances the detection of rare alleles and dominance effects, and allows the construction of complex genetic networks for traits like yield, using datasets from rice, maize, and soybeans.

In rice studies, Fast3VmrMLM identified 211 known genes and 384 multi-omics-supported candidate genes across 14 traits. It also uncovered 21 "hub" genes critical to genetic improvement. In maize datasets, 50 key genes were detected. The framework is further expandable to haplotype, lncRNA, and structural variation data, making it adaptable for future pangenome research.

This breakthrough offers a sustainable solution to the rising computational demand in the age of big data and artificial intelligence and sets a strong foundation for Smart Breeding 5.0. The study was co-led by HZAU doctoral students Wang Jingtian, Chen Ying, and Zhao Miaomiao, and supported by the National Natural Science Foundation of China.

Theoretical framework of the Fast3VmrMLM algorithm. [Photo/news.hzau.edu.cn]