Machine Learning Algorithms Applied to Identify Microbial Species by Their Motility.

Area of Science:

• Astrobiology
• Microbiology
• Machine Learning

Background:

• Future space missions require biochemistry-independent life detection methods.
• Advancements in computational power enable automated in-situ analysis of microscopic observations.
• Distinguishing alien life from abiotic processes is crucial for planetary exploration.

Purpose of the Study:

• To develop and test a semi-automated system for differentiating microbial motility from Brownian motion.
• To utilize machine learning for identifying specific bacterial species based on their movement patterns.
• To assess the feasibility of motility as a biosignature for extraterrestrial life detection.

Main Methods:

• A semi-automated experimental setup was designed to analyze microscopic movements.
• Supervised machine learning algorithms were trained to recognize bacterial motility.
• The system was tested against abiotic particles exhibiting Brownian motion and four bacterial species (P. haloplanktis, P. halocryophilus, B. subtilis, E. coli).

Main Results:

• Microbial motility was distinguished from Brownian motion with over 99% accuracy.
• Automated identification of specific bacterial species achieved an accuracy not exceeding 82%.
• Motility was confirmed as a reliable indicator of biological activity.

Conclusions:

• Motility serves as a robust biosignature for life-detection missions.
• The developed system provides a foundation for microscopic life recognition systems for missions to Mars and ocean worlds.
• Further refinement of machine learning algorithms is needed to improve species-specific identification accuracy.