Deep Video Analysis for Bacteria Genotype Prediction.

Area of Science:

• Microbiology
• Biotechnology
• Machine Learning

Background:

• Genetic modification of microbes is essential for industrial microbiology, bioproduction, and drug discovery.
• Linking specific genetic changes to bacterial phenotypes is critical for scientific advancement.

Purpose of the Study:

• To develop a supervised model for classifying bacteria with single gene modifications.
• To establish connections between bacterial genotype and observable phenotype.

Main Methods:

• Utilized low-resolution bright-field microscopy videos of *Vibrio cholerae* growth.
• Applied a supervised machine learning model for genotype classification.
• Introduced a weakly supervised approach to identify critical time points in bacterial growth.

Main Results:

• Spatiotemporal growth patterns in microscopy videos accurately predict bacterial genotype.
• Identified key moments in culture development that drive prediction accuracy.
• Demonstrated the link between specific gene modifications and observed phenotypes.

Conclusions:

• Machine learning can effectively classify bacterial genotypes based on phenotypic data from microscopy.
• This method offers insights into gene function and developmental stages.
• Automated phenotype analysis using machine learning has potential applications in drug discovery and disease management.