ChatGPT-based meta-analysis for evaluating the temporal and spatial characteristics of deoxynivalenol contamination in Chinese wheat

Abstract

Deoxynivalenol (DON) is a major source of mycotoxins in wheat. However, there is a lack of systematic reporting of the overall contamination status in China, hindering comprehensive assessments. In this study, we utilized a meta-analysis approach based on ChatGPT to systematically analyze DON contamination in wheat-growing regions in China, as reported in the literature from 2010 to 2021. By optimizing the query processes and refining the methodology keywords using ChatGPT, efficient screening, data identification, and literature extraction were achieved for the first time during the meta-analysis data acquisition phase. The matching rates for the screening and extraction of 1,091 articles were 100% and 95.4%, respectively, resulting in a 20.5-fold work efficiency increase compared to that by manual operations. Meta-subgroup analysis by province and year revealed significant spatiotemporal heterogeneity in DON contamination in the wheat-growing regions of China. Furthermore, the relationship between climate factors and DON levels in wheat was investigated to illustrate the spatial and temporal heterogeneity of DON in Chinese wheat. The results showed that DON concentrations were mainly influenced by relative humidity and precipitation during the wheat-growing season. This novel ChatGPT-assisted meta-analysis approach provides valuable insights and offers a promising method for efficient meta-analyses in other fields.