MYCOLab Ai Logo - gold and army green

Automatic classification of fungal-fungal interactions using deep learning models

Summary

Researchers developed a computer artificial intelligence system that can automatically analyze images of fungal interactions to identify strains that could help control harmful crop diseases. Instead of having humans manually examine thousands of fungal culture plates—a slow and subjective process—the AI system can now classify the interactions between beneficial fungi and plant pathogens with 95% accuracy. This breakthrough significantly speeds up the search for natural alternatives to synthetic pesticides, supporting sustainable agriculture and food security.

Background

Fungi are valuable sources of bioactive metabolites and enzymes for biotechnological applications including biocontrol organisms in agriculture. Current workflows for identifying biocontrol fungi rely on subjective visual observations of strain performance in microbe-microbe interaction studies, making the process time-consuming and difficult to reproduce.

Objective

To develop an AI-automated image classification approach using deep neural networks to analyze standardized images of 96-well microtiter plates with fungal-fungal challenge experiments. The study aimed to classify the outcome of interactions between the plant pathogen Fusarium graminearum and individual isolates from a collection of 38,400 fungal strains from the IBT collection.

Results

DenseNet121 achieved the highest performance with 95.0% accuracy, 92.4% precision, 94.1% recall, and 93.1% macro-averaged F1-score. ResNet50 also performed well with 94.9% accuracy. All five models demonstrated robust performance across multiple evaluation metrics, with DenseNet121 showing superior precision and consistency in capturing complex interaction patterns.

Conclusion

This study presents the first automated method for classifying fungal-fungal interactions using deep learning, successfully automating a previously subjective and time-consuming process. The high accuracy and precision of the DenseNet121 model demonstrate its potential as a valuable tool for efficient biocontrol organism identification, significantly advancing the field of mycology and sustainable agriculture.
  • Published in:Computational and Structural Biotechnology Journal,
  • Study Type:Experimental Study,
  • Source: PMID: 39655263, DOI: 10.1016/j.csbj.2024.11.027
Scroll to Top