MYCOLab Ai Logo - gold and army green

Identification of Burkholderia gladioli pv. cocovenenans in Black Fungus and Efficient Recognition of Bongkrekic Acid and Toxoflavin Producing Phenotype by Back Propagation Neural Network

Summary

This research investigated dangerous bacterial contamination in black fungus, a popular edible mushroom, and developed an innovative AI-based method to quickly identify toxic strains. The findings have significant implications for food safety and public health. Key impacts on everyday life: • Helps ensure safer black fungus products for consumers • Enables faster detection of dangerous contamination in food • Reduces the risk of severe food poisoning incidents • Provides new tools for food safety inspection and quality control • Demonstrates how AI can improve food safety testing

Background

Black fungus (Auricularia auricula-judae) is one of the four major cultivated edible fungi worldwide, with China producing 6.90-8.29 megatons annually from 2016-2020. However, contamination by Burkholderia gladioli pv. cocovenenans poses a serious food safety risk due to its production of deadly toxins bongkrekic acid and toxoflavin. Previous food poisoning incidents have affected over 9,000 people and caused more than 1,000 deaths in China alone.

Objective

To develop an efficient method for recognizing toxin-producing B. gladioli strains through exploring the potential of multilocus sequence typing and back propagation neural network for identification of toxigenic B. cocovenenans. The study aimed to isolate and characterize virulent strains from black fungus cultivation environments and evaluate their toxin production capabilities under different conditions.

Results

Twenty-six B. gladioli strains were isolated, with 84.6% being toxigenic. The isolates produced bongkrekic acid ranging from 0.05-6.24 mg/L in black fungus, with one highly toxigenic strain (NC18) generating 201.86 mg/L bongkrekic acid and 45.26 mg/L toxoflavin when co-cultivated with Rhizopus oryzae. The developed BP neural network model achieved 100% accuracy in training sets and 86.7% accuracy in external validation for predicting toxigenic phenotypes.

Conclusion

The study revealed widespread contamination of black fungus by B. gladioli pv. cocovenenans and demonstrated that MLST sequence data combined with BP neural network analysis provides an efficient method for identifying toxigenic strains without requiring complex cultivation and toxin determination procedures. This approach offers a rapid and convenient tool for pathogen detection and food safety surveillance.
  • Published in:Foods,
  • Study Type:Laboratory Research,
  • Source: 10.3390/foods13020351
Scroll to Top