Scanning electron microscopy and machine learning reveal heterogeneity in capsular morphotypes of the human pathogen...

Area of Science:

• Microbiology
• Computational Biology
• Mycology

Background:

• Phenotypic heterogeneity is crucial for microorganism survival, particularly in pathogens like Cryptococcus spp.
• Cryptococcus spp. are a significant global health threat, necessitating detailed characterization.

Purpose of the Study:

• To define and classify distinct Cryptococcus spp. capsule morphotypes.
• To develop an automated computational pipeline for identifying and categorizing these morphotypes.
• To explore potential associations between capsule morphology and yeast virulence.

Main Methods:

• Application of scanning electron microscopy (SEM) to visualize Cryptococcus spp. capsules.
• Development of an automated image analysis pipeline for object detection and feature computation.
• Utilizing a Random Forest (RF) machine learning classifier for morphotype classification.

Main Results:

• Four capsule morphotypes (Regular, Spiky, Bald, Phantom) were identified and consistently observed.
• The automated pipeline achieved 85% overall accuracy in classifying morphotypes.
• Structural and texture features were identified as key discriminators for classification.

Conclusions:

• The study successfully defined and computationally classified Cryptococcus spp. capsule morphotypes.
• The developed automated pipeline offers a robust method for rapid identification of morphological variants.
• This work lays the foundation for investigating the link between capsule morphology and Cryptococcus spp. virulence.