High-throughput microbial culturomics using automation and machine learning

Area of Science:

• Microbiome research
• Synthetic biology
• Genomics

Background:

• Isolating pure bacterial cultures is crucial for microbiome research but traditional methods are inefficient and lack integration.
• Current techniques struggle with scalability and combining phenotypic and genotypic data.

Purpose of the Study:

• To develop an open-source, high-throughput robotic platform for rapid, on-demand bacterial strain isolation.
• To integrate machine learning with colony morphology and genomic data for maximizing microbial diversity and enabling targeted isolation.
• To create personalized gut microbiome biobanks for comprehensive analysis.

Main Methods:

• An open-source robotic platform was engineered for high-throughput strain isolation.
• A machine learning model was developed using colony morphology and genomic data for microbe selection.
• The platform was applied to human fecal samples, generating extensive isolate collections and genomic data.

Main Results:

• Generated personalized gut microbiome biobanks from 20 human fecal samples, yielding 26,997 isolates representing over 80% of abundant taxa.
• Spatial analysis of over 100,000 colonies revealed cogrowth patterns suggesting significant microbial interactions.
• Comparative genomic analysis of 1,197 isolates uncovered strain evolution, selection, and horizontal gene transfer.

Conclusions:

• The culturomics framework enables systematic, quantitative analysis of imaging-based phenotypes and high-resolution genomics.
• This platform significantly advances the ability to generate and study diverse bacterial isolates for microbiome research.
• The findings provide insights into gut microbial interactions, evolution, and gene transfer.