Predicting invasive fungal pathogens using invasive pest assemblages: testing model predictions in a virtual world.

Area of Science:

• Ecology
• Computational Biology
• Biosecurity

Background:

• Predicting species invasions is challenging due to vast datasets.
• Existing methods struggle to analyze large numbers of potential invasive species simultaneously.
• Self-organizing maps (SOMs) offer a potential solution for ranking species by establishment likelihood.

Purpose of the Study:

• To assess the effectiveness of self-organizing maps (SOMs) for predicting species invasions.
• To validate the SOM method using a simulated invasive species dataset.
• To determine the reliability of SOM analysis for biosecurity risk assessment.

Main Methods:

• Applied a self-organizing map (SOM) to analyze the global distribution of 486 fungal pathogens.
• Developed a novel validation approach using a virtual world of invasive species to test SOM performance.
• Combined species richness data with SOM clustering patterns to refine predictions in sparsely populated regions.

Main Results:

• The SOM demonstrated high overall effectiveness, with an average success rate of 96-98% in the virtual world.
• Predictions were less accurate in regions with low species diversity (1-10 species).
• Reliability scores for fungal pathogen invasion risk in Australia ranged from 84-98%.

Conclusions:

• Self-organizing map analysis is a reliable method for assessing invasion risk from large datasets.
• SOMs can assist biosecurity agencies worldwide in prioritizing management efforts.
• This approach enhances the overall assessment of potential biological invasions.