Professor Liu Zhendong from the School of Computer and Information Engineering, in collaboration with the team of Professor Xiao Chuanle from the State Key Laboratory of Ophthalmology at Sun Yat-sen University, has published a paper titled DeepPlant: Accurate Cross-Species 5mC Detection for Oxford Nanopore Sequencing in Plants in Nature Communications (a top journal of the Chinese Academy of Sciences with an impact factor of 14.7 in 2024), a sub-journal of Nature, after years of research. This represents the first academic achievement published by our school in an international top journal and a sub-journal of Nature.

The paper presents the development of DeepPlant, a deep learning model that integrates the bidirectional long short-term memory network (Bi-LSTM) and Transformer architecture. This model significantly improves the detection accuracy of CHH sites and constructs a training-testing dataset covering diverse 9-mer motifs, effectively addressing the scarcity of positive training samples for CHH methylation. In evaluations across nine species, DeepPlant demonstrated a correlation of 0.705–0.838 between genome-wide methylation frequencies at CHH sites and bisulfite sequencing data, representing a 23.4%–117.6% improvement over Dorado. The model also exhibits excellent single-molecule detection accuracy and F1 scores, providing a powerful generalization-capable analytical tool for plant epigenetics research.