A team led by Professor Ye Kai at Xi'an Jiaotong University has introduced a groundbreaking genome language model, ANNEVO (Figure 1), which leverages a hybrid expert architecture. This model is capable of simultaneously learning evolutionary patterns across different biological taxa and understanding long-range sequence context relationships. Remarkably, ANNEVO enables high-precision de novo gene annotation based solely on DNA sequences, eliminating the need for external evidence such as RNA sequencing and homologous proteins.

Figure 1. Overview of the ANNEVO Method

The findings were published online on March 12, 2026, in the prestigious journal Nature Methods, under the title "High-precision de novo gene annotation using ANNEVO." Gene annotation is a crucial link between "genome sequencing" and "genome understanding," forming the foundation for advancing genome research towards functional analysis and practical applications. As large-scale international genome projects continue to generate vast amounts of data, achieving high-quality gene annotation has become a critical bottleneck in the post-genome era.

Traditional methods often depend on external evidence, which can be hampered by high data demands, significant computational costs, and limited applicability to species with scarce data. Professor Ye's team has adeptly addressed these challenges by integrating artificial intelligence with biomedicine. The team has long focused on "AI-driven genomic analysis" and has previously developed methods such as SVision (Nature Methods, 2022), SVision-pro (Nature Biotechnology, 2024), and Swave (Nature Genetics, 2026).

The introduction of ANNEVO further strengthens the team's comprehensive approach to "AI-driven genome analysis" (Figure 2), establishing a continuous methodological chain from genome variation identification to gene functional annotation. ANNEVO has shown substantial application potential in leading international genome projects, including the Darwin Tree of Life.

Figure 2. Overview of the team's series of achievements

The announcement of this achievement coincides with Xi'an Jiaotong University's anniversary and marks the tenth anniversary of Professor Ye Kai's return to the university, adding special significance to the milestone. This breakthrough highlights the team's decade-long exploration of the interdisciplinary frontier of "Artificial Intelligence + Genomics" and exemplifies the university's commitment to strengthening basic research, fostering interdisciplinary collaboration, and addressing the nation's major needs.

The research was conducted in close collaboration with Professor Meng Deyu's team from the School of Mathematics, showcasing the university's success in promoting interdisciplinary integration. Zhang Yizhe, an undergraduate from Xi'an Jiaotong University's Everest Program, actively contributed to the research, demonstrating the university's dedication to nurturing top-tier innovative talent. This achievement also signifies substantial progress in the university's ongoing promotion of the integration of "Artificial Intelligence + Life Sciences."

Zhang Pengyu, a doctoral student from the School of Automation, Department of Telecommunications, is the first author of the paper, with Professor Ye Kai serving as the corresponding author. The research received funding from key research and development projects by the Ministry of Science and Technology, the National Science Fund for Distinguished Young Scholars, and key projects from the National Natural Science Foundation of China.